Woodcut engraving element detection method and system

Abstract

The present invention discloses a method and system for detecting woodcut engraving elements, the method comprising: obtaining an image including a plurality of woodcut engraving elements as a sample image; performing 3D modeling according to the sample image to form a plurality of three-dimensional models; performing illumination preprocessing on the three-dimensional model, and performing image sampling to generate an enhanced sample image; annotating the enhanced sample image to generate a sample set, and training a deep neural network model according to the sample set to form a recognition model; and performing woodcut engraving element detection on the image to be detected by the recognition model. The present invention discloses a method and system for detecting woodcut engraving elements, which can effectively solve the sample deviation problem caused by defects and uneven illumination in existing Tibetan woodcut engravings through the above technical scheme, effectively improve the detection accuracy of the trained model in subsequent use, and at the same time enrich the sample amount that can be used for training, providing sufficient data for the research of Tibetan woodcut engravings.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method and system for detecting woodcut engraving elements. Background Art

[0002] Tibetan woodcuts are living fossils of Tibetan culture, carrying the wisdom of the Tibetan people. Studying Tibetan woodcuts can help us better understand and respect Tibetan culture and contribute to the protection and inheritance of intangible cultural heritage. Due to the long history and improper storage methods, woodcuts have been worn or aged naturally, so that subsequent researchers and cultural enthusiasts have encountered certain obstacles in understanding the elements of woodcuts. The development of deep learning and computer vision has provided an effective solution for the digital protection and inheritance of woodcut skills. As an important research direction in the field of computer vision, object detection uses deep learning algorithms to automatically identify and locate image and video content. Object detection algorithms have been widely used in various fields such as autonomous driving, biometric recognition, face recognition, and medical image analysis because of their advantages such as high efficiency and speed. Object detection algorithms can be divided into two categories: one is algorithms that rely on traditional technologies, and the other is algorithms that use deep learning technologies.

[0003] In the existing technology, no matter which target detection algorithm is used, a large number of high-definition samples are required to train the model in order to obtain a model with the expected accuracy. However, the sample defects provided by Tibetan woodcut samples due to wear and natural aging and different shooting lighting conditions will have a negative impact on the training of the detection model. Summary of the invention

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, the purpose of the present application is to provide a method and system for detecting woodcut engraving elements.

[0005] In a first aspect, an embodiment of the present application provides a method for detecting woodcut engraving elements, comprising:

[0006] Acquire an image including a plurality of woodcut engraving elements as a sample image;

[0007] Performing 3D modeling according to the sample image to form multiple three-dimensional models;

[0008] Performing illumination preprocessing on the three-dimensional model and performing image sampling to generate an enhanced sample image;

[0009] The enhanced sample image is annotated to generate a sample set, and a deep neural network model is trained according to the sample set to form a recognition model; the input data of the recognition model is the image data, and the output data of the recognition model is the type of woodcut element corresponding to the image data;

[0010] The woodcut elements are detected on the image to be detected by using the recognition model.

[0011] When implementing the embodiment of the present application, it is necessary to first collect sample images, which is mainly done by collecting images of existing Tibetan woodcuts of various styles. In order to obtain a large number of images for model training, a 3D modeling method is used in the embodiment of the present application to generate a three-dimensional model. After the three-dimensional model is generated, the defects and different lighting conditions of the original Tibetan woodcut can be corrected, which can effectively improve the accuracy of subsequent sampling samples.

[0012] In the embodiment of the present application, in order to provide richer samples, it is necessary to consider that the photographed woodcut engraving may be in different lighting conditions during the image recognition process, so it is necessary to perform illumination preprocessing to obtain richer sample information in the image sampling. Based on these enhanced sample images, manual annotation can be performed to generate a sample set. It should be understood that the enhanced sample image can be added to the original sample image to further enrich the sample, which should be within the protection scope of the embodiment of the present application. When training the deep neural network model, YOLOv5s is preferred. The process of training the recognition model and the operation process of the deep neural network model belong to the prior art, and the embodiment of the present application does not make much limitation. Finally, the trained recognition model can detect the elements of the woodcut engraving on the image to be detected. The embodiment of the present application can effectively solve the sample deviation problem caused by defects and uneven illumination in the existing Tibetan woodcut engravings through the above-mentioned technical solution, effectively improve the detection accuracy of the trained model in subsequent use, and enrich the sample volume that can be used for training, and provide sufficient data for the study of Tibetan woodcut engravings.

[0013] In a possible implementation, performing 3D modeling according to the sample image to form a plurality of three-dimensional models includes:

[0014] Performing edge detection on the sample image to form an edge point set and an intermediate point set; the edge point set is the points on the edge, and the intermediate point set is the points on the non-edge;

[0015] Assigning depth data to points of the edge point set and the intermediate point set to form a basic point set;

[0016] Performing surface fitting according to the points in the basic point set to form a continuous surface on the upper surface of the 3D model;

[0017] Select possible materials and color schemes according to the woodcut elements corresponding to the sample image;

[0018] The continuous curved surface is materialized and colored according to the material and color scheme to form a plurality of the three-dimensional models.

[0019] In a possible implementation, assigning depth data to points of the edge point set and the intermediate point set to form a basic point set includes:

[0020] According to the points in the edge point set, the points in the intermediate point set are identified as convex parts and concave parts to form convex points and concave points;

[0021] Assigning depth data using points in the edge point set as reference points;

[0022] Assigning depth data using the salient point corresponding to the middle line of the salient point clustering area as the highest point;

[0023] The depth data is assigned by taking the sunken point corresponding to the middle line of the sunken point gathering area as the lowest point; the depth data of the highest point, the reference point and the lowest point decrease in sequence.

[0024] In a possible implementation, performing illumination preprocessing on the three-dimensional model and performing image sampling to generate an enhanced sample image includes:

[0025] Setting light sources with multiple lighting angles and lighting intensities for the three-dimensional model;

[0026] Image sampling is performed on the three-dimensional model from the top under illumination of different light sources to form a plurality of enhanced sample images corresponding to the three-dimensional model.

[0027] In a possible implementation manner, annotating the enhanced sample image to generate a sample set includes:

[0028] The image containing at least two woodcut engraving elements is segmented, and the segmented enhanced sample image and the enhanced sample image that does not need to be segmented are annotated to generate a sample set; the annotated labels include graphic symbol category, deity category, folk custom category, text category, mantra wheel category and ritual instrument category.

[0029] In a possible implementation, training a deep neural network model according to the sample set to form a recognition model includes:

[0030] Splitting the sample set into a training set, a validation set, and a test set;

[0031] Using the image information in the training set as input data and the annotation labels corresponding to the image information as output information to train the deep neural network model to form the recognition model;

[0032] The recognition model is verified and tested through the verification set and the test set, and optimization is completed to form the final recognition model.

[0033] In a second aspect, the embodiment of the present application further provides a woodcut engraving element detection system, comprising:

[0034] An acquisition unit configured to acquire an image including a plurality of woodcut engraving elements as a sample image;

[0035] A modeling unit, configured to perform 3D modeling according to the sample image to form a plurality of three-dimensional models;

[0036] A preprocessing unit, configured to perform various preprocessing operations on the three-dimensional model and perform image sampling to generate an enhanced sample image;

[0037] A training unit is configured to annotate the enhanced sample image to generate a sample set, and train a deep neural network model according to the sample set to form a recognition model; the input data of the recognition model is the image data, and the output data of the recognition model is the type of woodcut element corresponding to the image data;

[0038] The detection unit is configured to perform woodcut engraving element detection on the image to be detected through the recognition model.

[0039] In a possible implementation, the modeling unit is further configured to:

[0040] Performing edge detection on the sample image to form an edge point set and an intermediate point set; the edge point set is the points on the edge, and the intermediate point set is the points on the non-edge;

[0041] Assigning depth data to points of the edge point set and the intermediate point set to form a basic point set;

[0042] Performing surface fitting according to the points in the basic point set to form a continuous surface on the upper surface of the 3D model;

[0043] Select possible materials and color schemes according to the woodcut elements corresponding to the sample image;

[0044] The continuous curved surface is materialized and colored according to the material and color scheme to form a plurality of the three-dimensional models.

[0045] In a possible implementation, the modeling unit is further configured to:

[0046] According to the points in the edge point set, the points in the intermediate point set are identified as convex parts and concave parts to form convex points and concave points;

[0047] Assigning depth data using points in the edge point set as reference points;

[0048] Assigning depth data using the salient point corresponding to the middle line of the salient point clustering area as the highest point;

[0049] The depth data is assigned by taking the sunken point corresponding to the middle line of the sunken point gathering area as the lowest point; the depth data of the highest point, the reference point and the lowest point decrease in sequence.

[0050] In a possible implementation manner, the preprocessing unit is further configured to:

[0051] Setting light sources with multiple lighting angles and lighting intensities for the three-dimensional model;

[0052] Image sampling is performed on the three-dimensional model from the top under illumination of different light sources to form a plurality of enhanced sample images corresponding to the three-dimensional model.

[0053] The present invention is achieved through the following technical solutions:

[0054] A method and system for detecting woodcut engraving elements, comprising:

[0055] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0056] The present invention provides a method and system for detecting woodcut elements. Through the above technical scheme, the sample deviation problem caused by defects and uneven lighting in existing Tibetan woodcuts can be effectively solved, and the detection accuracy of the trained model in subsequent use can be effectively improved. At the same time, the sample size that can be used for training is enriched, providing sufficient data for the research of Tibetan woodcuts. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0058] Figure 1 This is a schematic diagram of the method steps of an embodiment of the present application. DETAILED DESCRIPTION

[0059] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the drawings in the embodiment of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the embodiment of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0060] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0061] Please refer to Figure 1 , is a flow chart of a woodcut engraving element detection method provided in an embodiment of the present invention. Further, the woodcut engraving element detection method may specifically include the contents described in the following steps S1 to S5.

[0062] S1: Acquire an image including a variety of woodcut engraving elements as a sample image;

[0063] S2: performing 3D modeling according to the sample image to form multiple three-dimensional models;

[0064] S3: performing illumination preprocessing on the three-dimensional model, and performing image sampling to generate an enhanced sample image;

[0065] S4: annotating the enhanced sample image to generate a sample set, and training a deep neural network model according to the sample set to form a recognition model; the input data of the recognition model is the image data, and the output data of the recognition model is the type of woodcut element corresponding to the image data;

[0066] S5: Perform woodcut elements detection on the image to be detected by using the recognition model.

[0067] When implementing the embodiment of the present application, it is necessary to first collect sample images, which is mainly done by collecting images of existing Tibetan woodcuts of various styles. In order to obtain a large number of images for model training, a 3D modeling method is used in the embodiment of the present application to generate a three-dimensional model. After the three-dimensional model is generated, the defects and different lighting conditions of the original Tibetan woodcut can be corrected, which can effectively improve the accuracy of subsequent sampling samples.

[0068] In the embodiment of the present application, in order to provide richer samples, it is necessary to consider that the photographed woodcut engraving may be in different lighting conditions during the image recognition process, so it is necessary to perform illumination preprocessing to obtain richer sample information in the image sampling. Based on these enhanced sample images, manual annotation can be performed to generate a sample set. It should be understood that the enhanced sample image can be added to the original sample image to further enrich the sample, which should be within the protection scope of the embodiment of the present application. When training the deep neural network model, YOLOv5s is preferred. The process of training the recognition model and the operation process of the deep neural network model belong to the prior art, and the embodiment of the present application does not make much limitation. Finally, the trained recognition model can detect the elements of the woodcut engraving on the image to be detected. The embodiment of the present application can effectively solve the sample deviation problem caused by defects and uneven illumination in the existing Tibetan woodcut engravings through the above-mentioned technical solution, effectively improve the detection accuracy of the trained model in subsequent use, and enrich the sample volume that can be used for training, and provide sufficient data for the study of Tibetan woodcut engravings.

[0069] In a possible implementation, performing 3D modeling according to the sample image to form a plurality of three-dimensional models includes:

[0070] Performing edge detection on the sample image to form an edge point set and an intermediate point set; the edge point set is the points on the edge, and the intermediate point set is the points on the non-edge;

[0071] Assigning depth data to points of the edge point set and the intermediate point set to form a basic point set;

[0072] Performing surface fitting according to the points in the basic point set to form a continuous surface on the upper surface of the 3D model;

[0073] Select possible materials and color schemes according to the woodcut elements corresponding to the sample image;

[0074] The continuous curved surface is materialized and colored according to the material and color scheme to form a plurality of the three-dimensional models.

[0075] When the embodiment of the present application is implemented, in order to improve the modeling efficiency, it is necessary to perform edge detection on the sample image in the embodiment of the present application. It should be understood that the sample image is generally preprocessed before edge detection, such as graying and other processes, and the embodiment of the present application does not limit it. After edge detection, the corresponding pixels of the sample image are identified as edge points and non-edge points, wherein the edge points constitute an edge point set, and the non-edge points constitute an intermediate point set. At this time, the depth of these points can be assigned according to the relative position of the edge points and the non-edge points. The points with completed depth assignment can be used for surface fitting in the software to form a continuous surface, which can characterize the three-dimensional features of the sample image. When assigning values ​​to the continuous surface, it is necessary to match different materials and color schemes, wherein the material is mainly aimed at different reflective parameters, and the color can be selected according to the actual situation of the Tibetan woodcut engraving. The embodiment of the present application does not limit it, and those skilled in the art can make corresponding choices according to their needs.

[0076] In a possible implementation, assigning depth data to points of the edge point set and the intermediate point set to form a basic point set includes:

[0077] According to the points in the edge point set, the points in the intermediate point set are identified as convex parts and concave parts to form convex points and concave points;

[0078] Assigning depth data using points in the edge point set as reference points;

[0079] Assigning depth data using the salient point corresponding to the middle line of the salient point clustering area as the highest point;

[0080] The depth data is assigned by taking the sunken point corresponding to the middle line of the sunken point gathering area as the lowest point; the depth data of the highest point, the reference point and the lowest point decrease in sequence.

[0081] When the embodiment of the present application is implemented, a specific technical solution for depth data assignment is provided, wherein for edge points, the assigned value is a reference value, such as 0mm; and for convex points, the middle line is selected as the highest point, which can facilitate subsequent fitting, such as assigning a value of 5mm, and similarly for concave points, the middle line is selected as the lowest point, which can also facilitate subsequent fitting, such as assigning a value of -5mm. It should be understood that the middle line described in the embodiment of the present application can be selected according to the needs of those skilled in the art, such as selecting the middlemost line segment for a strip area, and selecting the line segment at the circle for a circular area, and the embodiment of the present application does not make much limitation.

[0082] In a possible implementation, performing illumination preprocessing on the three-dimensional model and performing image sampling to generate an enhanced sample image includes:

[0083] Setting light sources with multiple lighting angles and lighting intensities for the three-dimensional model;

[0084] Image sampling is performed on the three-dimensional model from the top under illumination of different light sources to form a plurality of enhanced sample images corresponding to the three-dimensional model.

[0085] When the embodiment of the present application is implemented, in the generation process of the enhanced sample image, the number of samples can be greatly enriched by selecting light sources with different angles and intensities.

[0086] In a possible implementation manner, annotating the enhanced sample image to generate a sample set includes:

[0087] The image containing at least two woodcut engraving elements is segmented, and the segmented enhanced sample image and the enhanced sample image that does not need to be segmented are annotated to generate a sample set; the annotated labels include graphic symbol category, deity category, folk custom category, text category, mantra wheel category and ritual instrument category.

[0088] When the embodiment of the present application is implemented, the woodcut engraving corresponding to the enhanced sample image may gather more than one type of elements. At this time, image segmentation is required to form a sample set. The LabelImg tool can be used for precise labeling and the labeling information can be saved in PascalVOC format.

[0089] In a possible implementation, training a deep neural network model according to the sample set to form a recognition model includes:

[0090] Splitting the sample set into a training set, a validation set, and a test set;

[0091] Using the image information in the training set as input data and the annotation labels corresponding to the image information as output information to train the deep neural network model to form the recognition model;

[0092] The recognition model is verified and tested through the verification set and the test set, and optimization is completed to form the final recognition model.

[0093] When the embodiment of the present application is implemented, the obtained XML format tags are converted into a txt format suitable for the YOLOv5s model, and the data set is randomly and automatically divided into a training set, a validation set, and a test set in an 8:1:1 ratio. At the same time, the feature extraction module in the preferred YOLOv5s model is replaced with FasterNet to solve the problems of large parameters and long running time of the original YOLOv5s, so that the detection model is more lightweight, and the CA attention mechanism is added after the feature extraction network, which can better utilize high-level feature representations and capture important feature information before global pooling, so that the edge parts and small targets of woodcut engravings can be better detected. The introduction of lightweight upsampling CARAFE instead of nearest neighbor interpolation upsampling can enable the model to obtain a larger receptive field and retain more information. The improved algorithm can improve the detection speed while improving the detection accuracy of the model.

[0094] Based on the same inventive concept, the embodiment of the present application also provides a woodcut engraving element detection system, including:

[0095] An acquisition unit configured to acquire an image including a plurality of woodcut engraving elements as a sample image;

[0096] A modeling unit, configured to perform 3D modeling according to the sample image to form a plurality of three-dimensional models;

[0097] A preprocessing unit, configured to perform various preprocessing operations on the three-dimensional model and perform image sampling to generate an enhanced sample image;

[0098] A training unit is configured to annotate the enhanced sample image to generate a sample set, and train a deep neural network model according to the sample set to form a recognition model; the input data of the recognition model is the image data, and the output data of the recognition model is the type of woodcut element corresponding to the image data;

[0099] The detection unit is configured to perform woodcut engraving element detection on the image to be detected through the recognition model.

[0100] In a possible implementation, the modeling unit is further configured to:

[0101] Performing edge detection on the sample image to form an edge point set and an intermediate point set; the edge point set is the points on the edge, and the intermediate point set is the points on the non-edge;

[0102] Assigning depth data to points of the edge point set and the intermediate point set to form a basic point set;

[0103] Performing surface fitting according to the points in the basic point set to form a continuous surface on the upper surface of the 3D model;

[0104] Select possible materials and color schemes according to the woodcut elements corresponding to the sample image;

[0105] The continuous curved surface is materialized and colored according to the material and color scheme to form a plurality of the three-dimensional models.

[0106] In a possible implementation, the modeling unit is further configured to:

[0107] According to the points in the edge point set, the points in the intermediate point set are identified as convex parts and concave parts to form convex points and concave points;

[0108] Assigning depth data using points in the edge point set as reference points;

[0109] Assigning depth data using the salient point corresponding to the middle line of the salient point clustering area as the highest point;

[0110] The depth data is assigned by taking the sunken point corresponding to the middle line of the sunken point gathering area as the lowest point; the depth data of the highest point, the reference point and the lowest point decrease in sequence.

[0111] In a possible implementation manner, the preprocessing unit is further configured to:

[0112] Setting light sources with multiple lighting angles and lighting intensities for the three-dimensional model;

[0113] Image sampling is performed on the three-dimensional model from the top under illumination of different light sources to form a plurality of enhanced sample images corresponding to the three-dimensional model.

[0114] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0115] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.

[0116] The units described as separate components may or may not be physically separated. As units, it is obvious that a person of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0117] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0118] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a grid device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0119] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting woodcut engraving elements, characterized in that: include: Acquire an image including a plurality of woodcut engraving elements as a sample image; Performing 3D modeling according to the sample image to form multiple three-dimensional models; Performing illumination preprocessing on the three-dimensional model and performing image sampling to generate an enhanced sample image; The enhanced sample image is annotated to generate a sample set, and a deep neural network model is trained according to the sample set to form a recognition model; the input data of the recognition model is the image data, and the output data of the recognition model is the type of woodcut element corresponding to the image data; Performing woodcut engraving element detection on the image to be detected by using the recognition model; Performing 3D modeling according to the sample image to form multiple three-dimensional models includes: Performing edge detection on the sample image to form an edge point set and an intermediate point set; the edge point set is the points on the edge, and the intermediate point set is the points on the non-edge; Assigning depth data to points of the edge point set and the intermediate point set to form a basic point set; Performing surface fitting according to the points in the basic point set to form a continuous surface on the upper surface of the 3D model; Select possible materials and color schemes according to the woodcut elements corresponding to the sample image; Arranging and coloring the continuous curved surface according to the material and color scheme to form a plurality of the three-dimensional models; Assigning depth data to the points of the edge point set and the intermediate point set to form a basic point set includes: According to the points in the edge point set, the points in the intermediate point set are identified as convex parts and concave parts to form convex points and concave points; Assigning depth data using points in the edge point set as reference points; Assigning depth data using the salient point corresponding to the middle line of the salient point clustering area as the highest point; The depth data is assigned by taking the sunken point corresponding to the middle line of the sunken point gathering area as the lowest point; the depth data of the highest point, the reference point and the lowest point decrease in sequence.

2. A method for detecting woodcut elements according to claim 1, characterized in that: Performing illumination preprocessing on the three-dimensional model and performing image sampling to generate enhanced sample images includes: Setting light sources with multiple lighting angles and lighting intensities for the three-dimensional model; Image sampling is performed on the three-dimensional model from the top under illumination of different light sources to form a plurality of enhanced sample images corresponding to the three-dimensional model.

3. A method for detecting woodcut elements according to claim 1, characterized in that: Annotating the enhanced sample image to generate a sample set includes: The image containing at least two woodcut engraving elements is segmented, and the segmented enhanced sample image and the enhanced sample image that does not need to be segmented are annotated to generate a sample set; the annotated labels include graphic symbol category, deity category, folk custom category, text category, mantra wheel category and ritual instrument category.

4. A method for detecting woodcut elements according to claim 1, characterized in that: Training a deep neural network model according to the sample set to form a recognition model includes: Splitting the sample set into a training set, a validation set, and a test set; Using the image information in the training set as input data and the annotation labels corresponding to the image information as output information to train the deep neural network model to form the recognition model; The recognition model is verified and tested through the verification set and the test set, and optimization is completed to form the final recognition model.

5. A woodcut engraving element detection system, characterized in that: include: An acquisition unit configured to acquire an image including a plurality of woodcut engraving elements as a sample image; A modeling unit, configured to perform 3D modeling according to the sample image to form a plurality of three-dimensional models; A preprocessing unit, configured to perform various preprocessing operations on the three-dimensional model and perform image sampling to generate an enhanced sample image; A training unit is configured to annotate the enhanced sample image to generate a sample set, and train a deep neural network model according to the sample set to form a recognition model; the input data of the recognition model is the image data, and the output data of the recognition model is the type of woodcut element corresponding to the image data; A detection unit, configured to detect woodcut elements on the image to be detected by using the recognition model; The modeling unit is further configured to: Performing edge detection on the sample image to form an edge point set and an intermediate point set; the edge point set is the points on the edge, and the intermediate point set is the points on the non-edge; Assigning depth data to points of the edge point set and the intermediate point set to form a basic point set; Performing surface fitting according to the points in the basic point set to form a continuous surface on the upper surface of the 3D model; Select possible materials and color schemes according to the woodcut elements corresponding to the sample image; Arranging and coloring the continuous curved surface according to the material and color scheme to form a plurality of the three-dimensional models; The modeling unit is further configured to: According to the points in the edge point set, the points in the intermediate point set are identified as convex parts and concave parts to form convex points and concave points; Assigning depth data using points in the edge point set as reference points; Assigning depth data using the salient point corresponding to the middle line of the salient point clustering area as the highest point; The depth data is assigned by taking the concave point corresponding to the middle line of the concave point gathering area as the lowest point; The depth data of the highest point, the reference point and the lowest point decrease in sequence.

6. A woodcut engraving element detection system according to claim 5, characterized in that: The pre-processing unit is further configured to: Setting light sources with multiple lighting angles and lighting intensities for the three-dimensional model; Image sampling is performed on the three-dimensional model from the top under illumination of different light sources to form a plurality of enhanced sample images corresponding to the three-dimensional model.

Images