Dynamic exhibition system for digital museum exhibits

Abstract

The invention discloses a dynamic exhibition system for digital museum exhibits, and relates to the field of exhibition design of exhibits, and the system is characterized in that an information collection module is used for collecting comprehensive original information, processing the comprehensive original information, and obtaining a comprehensive data set; the exhibition form generation module is used for constructing and training an exhibit enrichment model, performing input design on the comprehensive data set to generate preset data, and importing the preset data into the trained exhibit enrichment model to obtain an exhibit multi-form set; the optimization algorithm module is used for identifying an optimal exhibition form in an exhibit multi-form set through an optimization algorithm and displaying the exhibit in the optimal exhibition form, all the modules are connected in a wired and / or wireless mode, self-adaptive dynamic design of the exhibition mode of the exhibit in the digital museum is achieved, and the self-adaptive dynamic design of the exhibition mode of the exhibit in the digital museum is achieved. And the practicability of exhibit display is improved.

Description

Technical Field

[0001] The present invention relates to the field of exhibition design of exhibits, and more specifically, to a dynamic exhibition system of digital museum exhibits. Background Art

[0002] Traditional exhibit displays are relatively simple, usually static, lacking interactivity and diversity, and are difficult to attract the audience's continued attention, especially the younger generation. In addition, in museums or online exhibitions, due to the limitations of the display, the information conveyed by the exhibits is often not comprehensive enough, and the audience cannot fully understand the history, background and cultural significance of the exhibits, resulting in a reduction in the educational effect of the exhibition.

[0003] In addition, different audiences have different interests and needs. Traditional exhibition systems are rarely able to analyze user behavior in real time and dynamically adjust exhibition content based on the analysis results. This results in the exhibition experience being unable to adapt to changes in audience needs in a timely manner, and it is also impossible to effectively collect valuable user feedback. As a result, traditional exhibition forms often lack personalized displays and cannot be dynamically adjusted according to audience preferences. In the design and selection of exhibition forms, how to balance user experience, information communication effects, and resource utilization efficiency is a complex issue. Traditional methods usually rely on manual design, and the effects of exhibition forms are difficult to quantify and evaluate. Museum administrators and curators find it difficult to know the actual effects of a certain display form and the audience's acceptance, which affects the optimization and improvement of the exhibition, resulting in low exhibition efficiency and difficulty in achieving optimal results. Therefore, how to design dynamic exhibitions of digital museum exhibits is now an urgent issue to be solved. Summary of the invention

[0004] In view of the problems existing in the prior art, the object of the present invention is to provide a dynamic exhibition system for digital museum exhibits, so as to realize the adaptive dynamic design of the exhibition mode of exhibits in the digital museum.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a dynamic exhibition system for digital museum exhibits, the dynamic exhibition system for digital museum exhibits comprising:

[0006] Information collection module: used to collect and process comprehensive original information to obtain a comprehensive data set;

[0007] Exhibition form generation module: build and train an exhibit enrichment model, input the comprehensive data set to generate preset data, import the preset data into the trained exhibit enrichment model, and obtain a multi-form set of exhibits;

[0008] Optimization algorithm module: Use optimization algorithm to identify the best exhibition form among multiple forms of exhibits, and use the best exhibition form to display the exhibits;

[0009] The modules are connected to each other via wired and / or wireless means.

[0010] Preferably, the comprehensive original information includes basic information of exhibits, evaluation feedback information and preview information. The data categories of basic information of exhibits include background information, appearance information and multimedia information of exhibits. The data categories of evaluation feedback information include audience behavior data and expert value assessment data. The data categories of preview information include the audience and display environment of exhibits.

[0011] Preferably, the processing of the comprehensive original information to obtain the comprehensive data set specifically includes:

[0012] Identify the text in the comprehensive original information, pre-process the text and perform feature extraction to obtain a feature set;

[0013] Extract basic information of exhibits from comprehensive original information, and build a 3D simulation model based on the appearance information of exhibits;

[0014] The feature set is synaptically coupled with the three-dimensional simulation model to form a comprehensive data set.

[0015] Preferably, the method of obtaining a feature set after preprocessing the text and extracting features includes:

[0016] Assume that the text of each data category includes R_t documents. Assume that the vocabulary C_h = {w1,…,wj,…,wn} is the set of all words in the entire text. Use TF-IDF to convert any document di into a TF-IDF value represented by a word frequency vector, and then generate a feature vector V for each document. di =[C_p(w1,di),……,C_p(wn,di)], where the TF-IDF value of word wj in document di is calculated TF(wj,di) represents the frequency of word wj in document di, DF(wj) is the number of documents in which word wj appears, that is, the document frequency, and i=1, ..., R_t, j=1, ..., n, Among them, r_tij represents the number of occurrences of word wj in document di, and w_dj represents the total number of all words in document di;

[0017] Construct a P_Lasso regression model and set the objective function as Among them, β u_o is the regression coefficient of the u_oth TF-IDF value, γ is the regularization parameter used to control the sparsity of the model, and x w_o，u_ois the u_oth TF-IDF value of the w_oth word, W_F is the number of words, U_F is the number of TF-IDF values, and the component value of the w_oth word Among them, P(wj, Z_T) represents the joint probability of word wj and data category Z_Tb, that is, the probability of co-appearing in the document, P(wj) represents the marginal probability of word wj, that is, the probability of word wj appearing in the document, and P(Z_Tb) represents the marginal probability of the bth data category Z_Tb in the document;

[0018] Using the feature vector as input and the component value of the word as output, the value of the regression coefficient is adjusted by training the P_Lasso regression model, and the words with non-zero regression coefficients are retained to obtain the feature set.

[0019] Preferably, the method for preprocessing text includes:

[0020] Remove noise and stop words from the text, convert it into a unified format, and perform word segmentation after stem extraction and lemma restoration to obtain comprehensive original information after preprocessing.

[0021] Preferably, the method of synapsing the feature set with the three-dimensional simulation model comprises:

[0022] Identify each feature category data package in the feature set, match the feature category data package with the corresponding attributes of the 3D simulation model, set a click point at the corresponding location of the 3D simulation model match, hyperlink the click point to the corresponding feature category data package, add the feature category data package to the 3D simulation model of the exhibit, and form a comprehensive data set.

[0023] Preferably, the method of constructing and training the exhibit enrichment model includes:

[0024] A preset sample set, which includes o_p groups of data, each of which includes preset data of an exhibit and i_p exhibition forms corresponding to the exhibit;

[0025] The exhibit enrichment model is designed based on Text2Shape. The conditional cVAE architecture is used to design the generator. The input of the generator includes preset data, and the output is the exhibition form generated based on the input. The loss function is set to in, For all noise vectors P Z (Z), D(G(Z, F_text, x_3d), F_text) represents the authenticity score of the discriminator on the three-dimensional shape G(Z, F_text, x_3d) and the text description F_text in the generated exhibition form, x_3d represents the three-dimensional shape, and Z represents the noise vector;

[0026] The discriminator is designed to determine whether the generated exhibition form conforms to the input preset data. The input of the discriminator is the generated exhibition form. The loss function is set to in, Represents all real three-dimensional shapes P data The expected value of the true 3D shape x_3d in (x_3d), Represents all generated 3D shapes Producing 3D shapes The expected value of , D(x_3d, F_text) represents the authenticity score of the discriminator for the true 3D shape x_3d and the text description F_text, Represents the discriminator to generate a 3D shape and authenticity score of the text description F_text;

[0027] The learning rate is set, and the sample set is divided into multiple batches and input into the exhibit enrichment model in sequence until the expected effect is achieved and the training is stopped to obtain a trained exhibit enrichment model.

[0028] Preferably, the method of inputting the comprehensive data set to generate preset data includes:

[0029] For the text in the comprehensive dataset, a pre-trained language model is used to convert the text description into an embedding vector, which represents the semantic information of the text;

[0030] Use a 3D convolutional neural network or a point cloud-based neural network to encode the 3D simulation model, convert the appearance information of the 3D simulation model into a dimensional vector, and extract the key features of the model;

[0031] Collect the embedding vectors and dimension vectors of the exhibits to form preset data.

[0032] Preferably, the method of identifying the best exhibition form in a set of multiple exhibit forms by using an optimization algorithm comprises:

[0033] The optimization algorithm is designed based on Fish School Search, and the optimization goal is defined as maximizing the value of the evaluation function. The number of fish schools A_z and the position vector Wz_m are initialized, and the evaluation function is Pj_h=∈1×∑ zj_u∈pd zj_u+∈2×∑ gz_k∈xw gz_k+∈3×hj_qh+∈4×qh_d, where zj_u represents the evaluation value of the u-th data category in the expert's value evaluation data pd, gz_k represents the evaluation value of the k-th data category in the audience's behavior data xw, hj_qh represents the fit between the exhibition form and the exhibition environment, qh_d represents the degree of innovation of the exhibition form, ∈1, ∈2, ∈3, ∈4 represent weight coefficients respectively;

[0034] Set the foraging behavior: in, represents the position vector of the fth fish in the gth round, represents the position vector of the fth fish at the g-1th round, σ is the learning rate, if the value of the evaluation function of the fth fish at the g-1th round Less than the value of the evaluation function at round g Update for

[0035] The movement direction of the fth fish Among them, Wz_m f represents the position vector of the fth fish, Wz_m l represents the feature vector of the lth fish in the neighborhood Nb(f) of the fth fish, Pj_h(Wz_m f ) represents the position vector Wz_m of the fth fish f The value of the evaluation function at

[0036] Set the center of mass position of the fish school to be updated Among them, if the value of the evaluation function at the current centroid position is higher than the previous best position, the best position is updated, the number of iterations is set to U_KB, and it stops when the maximum number of iterations is reached. The current best position is extracted, and the exhibition form at this best position is set as the best exhibition form.

[0037] Preferably, the displaying of the exhibits using the best display format includes: arranging the best display format on the display terminal, and monitoring the running status of the display format in real time.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] In the present invention, through the information collection module, the background information, appearance information and multimedia information of the exhibits can be fully collected, and the information can be conveyed in combination with multimedia content such as text, audio, and video during the display process. Ensure that the relevant information of the exhibits can be fully displayed, help the audience to better understand and appreciate the exhibits, generate a variety of display forms based on the rich information of the exhibits through the exhibition form generation module and the multi-objective optimization algorithm, and customize the display mode of the exhibits according to the interests of tourists through user behavior analysis, so that the display of the exhibits is more diversified, interactive and innovative, and select the best display form through the optimization algorithm to find the best balance between user experience, information communication and resource utilization efficiency. This solution greatly improves the design efficiency and effect of the exhibition form, thereby improving the audience stickiness and satisfaction, and enhancing the audience's experience and participation; the adaptive control optimization mechanism continuously improves the overall efficiency of intelligent analysis, and multiple modules cooperate with each other to achieve comprehensive data collection, processing and analysis, and provide real-time and accurate analysis of detection data. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The present invention provides a structural schematic diagram of a dynamic exhibition system for digital museum exhibits;

[0041] Figure 2 The figure is a schematic diagram of the steps of the dynamic exhibition system applied to digital museum exhibits in the present invention. DETAILED DESCRIPTION

[0042] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0043] Embodiment 1

[0044] Reference Figure 1 and Figure 2 Embodiment 1 further illustrates a dynamic exhibition system for digital museum exhibits proposed by the present invention.

[0045] Traditional exhibit displays are relatively simple, usually static, lacking interactivity and diversity, and are difficult to attract the audience's continued attention, especially the younger generation. In addition, in museums or online exhibitions, due to the limitations of the display, the information conveyed by the exhibits is often not comprehensive enough, and the audience cannot fully understand the history, background and cultural significance of the exhibits, resulting in a reduction in the educational effect of the exhibition.

[0046] In addition, different audiences have different interests and needs. Traditional exhibition systems are rarely able to analyze user behavior in real time and dynamically adjust exhibition content based on the analysis results. This results in the exhibition experience being unable to adapt to changes in audience needs in a timely manner, and it is also impossible to effectively collect valuable user feedback. As a result, traditional exhibition forms often lack personalized displays and cannot be dynamically adjusted according to audience preferences. In the design and selection of exhibition forms, how to balance user experience, information communication effects, and resource utilization efficiency is a complex issue. Traditional methods usually rely on manual design, and the effects of exhibition forms are difficult to quantify and evaluate. Museum administrators and curators find it difficult to know the actual effects of a certain display form and the audience's acceptance, which affects the optimization and improvement of the exhibition, resulting in low efficiency in the display of exhibits and difficulty in achieving optimal results.

[0047] The present invention can comprehensively collect background information, appearance information and multimedia information of exhibits through the information collection module, and combine text, audio, video and other multimedia content to convey information during the display process. Ensure that the relevant information of the exhibits can be fully displayed to help the audience better understand and appreciate the exhibits. Through the exhibition form generation module and the multi-objective optimization algorithm, a variety of display forms are generated based on the rich information of the exhibits. Through user behavior analysis, the display mode of the exhibits is customized according to the interests of tourists, making the display of the exhibits more diversified, interactive and innovative, and the optimal display form is selected through the optimization algorithm to find the best balance between user experience, information communication and resource utilization efficiency. This solution greatly improves the design efficiency and effect of the exhibition form, thereby improving the audience stickiness and satisfaction, and enhancing the audience's experience and participation.

[0048] The system design of the present invention allows for rapid updating and expansion of exhibits and exhibition content. Administrators can easily upload new information and adjust exhibition formats to ensure the timeliness of exhibit content. At the same time, the architecture design based on cloud computing and content distribution network (CDN) ensures that the system can be quickly expanded to cope with high concurrent access requirements.

[0049] The system monitors and analyzes user interaction, dwell time and other data in real time through user behavior analysis. These data can not only be used to optimize the current exhibition format, but also dynamically adjust the exhibition content through the intelligent recommendation system to provide a more personalized experience that meets the needs of the audience. In addition, the system also provides a feedback mechanism that allows users to directly rate and comment on the exhibition format. These feedbacks will be used to continuously improve the exhibition content and format.

[0050] The present invention aims to generate multiple possible exhibition forms by collecting and analyzing all the information of the exhibits, select the best exhibition form by using the optimization algorithm, and finally generate the exhibition content displayed to the audience. The system is mainly divided into four main modules: information collection module, exhibition form generation module, optimization algorithm module, and display module, among which:

[0051] Information collection module: used to collect and process comprehensive original information to obtain a comprehensive data set;

[0052] The comprehensive original information includes basic information of exhibits, evaluation feedback information and preview information. The data categories of basic information of exhibits include background information, appearance information and multimedia information of exhibits. The data categories of evaluation feedback information include audience behavior data and expert value assessment data. The data categories of preview information include the audience and display environment of exhibits.

[0053] Background information may include the historical background, cultural significance, archaeological information, etc. of the exhibits; appearance information may include the size, material, color, shape, texture, etc. of the exhibits; multimedia information may include high-definition pictures, 3D models, videos, sound recordings, etc. of the exhibits.

[0054] Audience behavior data may include audience feedback on different exhibition formats (such as satisfaction surveys, comments, depth of understanding and learning, or the amount of exhibit content remembered), and audience behavior patterns in the exhibition (such as dwell time, interaction frequency); expert value assessment data may include experts' professional assessment of different exhibition formats, for example, the completeness of the content and the degree of completion of educational objectives.

[0055] The audience of the exhibits refers to the audience group that the exhibits are mainly aimed at (such as students, researchers, and the general public). The display environment includes the form of display, such as touch screen interaction, 3D models, virtual reality (VR), augmented reality (AR) display, etc., as well as the scope of the display.

[0056] The data sources of the information collection module include: museum databases, literature, archaeological reports, images and video materials, expert interviews, etc.; collecting opinions and feedback by distributing questionnaires to audiences, experts, and curators; recording audience behavior and interactive data during the exhibition; conducting in-depth interviews with the authors of the exhibits and relevant experts; conducting actual tests of virtual display technology, recording technical requirements and effects; analyzing the data and effects of previous similar exhibitions to extract valuable information.

[0057] Process the comprehensive original information to obtain a comprehensive data set, including:

[0058] Identify the text in the comprehensive original information, pre-process the text and perform feature extraction to obtain a feature set;

[0059] After preprocessing the text and performing feature extraction, the method of obtaining the feature set includes:

[0060] Remove noise and stop words from the text, convert it into a unified format, and perform word segmentation after stem extraction and lemma restoration to obtain comprehensive original information after preprocessing;

[0061] Noise usually refers to unnecessary characters, symbols, numbers, HTML tags, etc. in the text. The present invention removes noise through the HTML parser to make the text cleaner and reduce complexity. Stop words refer to words that are very common in the text but contribute less to the analysis task. The present invention can use the built-in stop word list in libraries such as NLTK and SpaCy to remove them. Stemming and lemmatization are two technologies for normalizing word forms. Stemming obtains the root by removing the affixes of the word, and lemmatization restores the word to its basic form (lemma). The present invention performs word segmentation processing through SpaCy Tokenizer.

[0062] Assume that the text of each data category includes R_t documents. Assume that the vocabulary C_h = {w1,…,wj,…,wn} is the set of all words in the entire text. Use TF-IDF to convert any document di into a TF-IDF value represented by a word frequency vector, and then generate a feature vector V for each document. di =[C_p(w1,di),……,C_p(wn,di)], where the TF-IDF value of word wj in document di is calculated TF(wj,di) represents the frequency of word wj in document di, DF(wj) is the number of documents in which word wj appears, that is, the document frequency, and i=1, ..., R_t, j=1, ..., n, Among them, r_tij represents the number of occurrences of word wj in document di, and w_dj represents the total number of all words in document di;

[0063] Construct a P_Lasso regression model and set the objective function as Among them, β u_o is the regression coefficient of the u_oth TF-IDF value, γ is the regularization parameter used to control the sparsity of the model, and x w_o，u_o is the u_oth TF-IDF value of the w_oth word, W_F is the number of words, U_F is the number of TF-IDF values, and the component value of the w_oth word Among them, P(wj, Z_T) represents the joint probability of word wj and data category Z_Tb, that is, the probability of co-appearing in the document, P(wj) represents the marginal probability of word wj, that is, the probability of word wj appearing in the document, and P(Z_Tb) represents the marginal probability of the bth data category Z_Tb in the document;

[0064] Using the feature vector as input and the component value of the word as output, the value of the regression coefficient is adjusted by training the P_Lasso regression model. In this process, the P_Lasso regression model will reduce the coefficients of some unimportant features (words) to zero, thereby retaining only important features and retaining features (words) with non-zero regression coefficients. These words are the feature words that have the greatest impact on the task, and obtain the feature set.

[0065] In high-dimensional text data, there are usually a large number of features, and not all of them contribute to the final task. To this end, Lasso regression (Least Absolute Shrinkage and Selection Operator) can be used for feature selection. In text feature extraction, Lasso regression can help identify and retain the words that are most relevant to the task (such as classification, sentiment analysis, etc.), while ignoring those less important words, making the model more efficient and easier to explain.

[0066] Extract basic information of exhibits from comprehensive original information, and build a 3D simulation model based on the appearance information of exhibits;

[0067] The 3D simulation model can be constructed through 3D scanning and modeling software.

[0068] Synapse the feature set with the three-dimensional simulation model to form a comprehensive data set;

[0069] Methods for synapsing the feature set with the three-dimensional simulation model include:

[0070] Identify each feature category data package in the feature set, match the feature category data package with the corresponding attribute of the three-dimensional simulation model, set a click point at the corresponding position of the three-dimensional simulation model, hyperlink the click point to the corresponding feature category data package, add the feature category data package to the three-dimensional simulation model of the exhibit, and form a comprehensive data set;

[0071] A three-dimensional display of exhibits and related information is achieved. When in use, the feature category data package with a hyperlink is expanded by starting the click point. The feature category data package generally includes a detailed description of the corresponding position of the exhibit or the feature category.

[0072] Exhibition form generation module: build and train an exhibit enrichment model, input the comprehensive data set to generate preset data, import the preset data into the trained exhibit enrichment model, and obtain a multi-form set of exhibits;

[0073] Ways to build and train an exhibit enrichment model include:

[0074] A preset sample set, which includes o_p groups of data, each of which includes preset data of an exhibit and i_p exhibition forms corresponding to the exhibit;

[0075] The exhibit enrichment model is designed based on Text2Shape. The conditional cVAE architecture is used to design the generator. The input of the generator includes preset data, and the output is the exhibition form generated based on the input. The loss function is set to in, For all noise vectors P Z (The expected value of Z(, D(G(Z, F_text, x_3d), F_text)) represents the authenticity score of the discriminator on the three-dimensional shape G(Z, F_text, x_3d) and the text description F_text in the generated exhibition form, x_3d represents the three-dimensional shape, and Z represents the noise vector;

[0076] The discriminator is designed to determine whether the generated exhibition form conforms to the input preset data. The input of the discriminator is the generated exhibition form. The loss function is set to in, Represents all real three-dimensional shapes P data The expected value of the true 3D shape x_3d in (x_3d), Represents all generated 3D shapes Producing 3D shapes The expected value of , D(x_3d, F_text) represents the authenticity score of the discriminator for the true 3D shape x_3d and the text description F_text, Represents the discriminator to generate a 3D shape and authenticity score of the text description F_text;

[0077] The learning rate is set, and the sample set is divided into multiple batches and input into the exhibit enrichment model in sequence until the expected effect is achieved and the training is stopped to obtain a trained exhibit enrichment model.

[0078] The expected effect can be set as multiple iterations, and the exhibit enrichment model training is stopped until the generator can stably produce high-quality three-dimensional shapes and the discriminator can effectively distinguish between true and false samples.

[0079] The methods of inputting the comprehensive data set to generate preset data include:

[0080] For the text in the comprehensive dataset, use a pre-trained language model (such as BERT, GPT-3, GloVe, etc.) to convert the text description into an embedding vector, which represents the semantic information of the text and serves as the input of the language model;

[0081] Use 3D convolutional neural networks (3D-CNN) or point cloud-based neural networks (such as PointNet) to encode the 3D simulation model, convert the appearance information of the 3D simulation model into a dimensional vector, and extract the key features of the model, such as shape, size, surface properties, etc. These features will be used to guide the generation process;

[0082] Collect the embedding vectors and dimension vectors of the exhibits to form preset data.

[0083] Optimization algorithm module: Use optimization algorithm to identify the best exhibition form among multiple forms of exhibits, and use the best exhibition form to display the exhibits;

[0084] Methods for identifying the best exhibition form in a multi-form set of exhibits through optimization algorithms include:

[0085] The optimization algorithm is designed based on Fish School Search, and the optimization goal is defined as maximizing the value of the evaluation function. The number of fish schools A_z and the position vector Wz_m are initialized, and the evaluation function is Pj_h=∈1×∑ zj_u∈pd zj_u+∈2×∑ gz_k∈xw gz_k+∈3×hj_qh+∈4×qh_d, where zj_u represents the evaluation value of the u-th data category in the expert's value evaluation data pd, gz_k represents the evaluation value of the k-th data category in the audience's behavior data xw, hj_qh represents the fit between the exhibition form and the exhibition environment, qh_d represents the degree of innovation of the exhibition form, ∈1, ∈2, ∈3, ∈4 represent weight coefficients respectively;

[0086] Set the foraging behavior: in, represents the position vector of the fth fish in the gth round, represents the position vector of the fth fish at the g-1th round, σ is the learning rate, if the value of the evaluation function of the fth fish at the g-1th round Less than the value of the evaluation function at round g Update for

[0087] The movement direction of the fth fish Among them, Wz_m f represents the position vector of the fth fish, Wz_m lrepresents the feature vector of the lth fish in the neighborhood Nb(f) of the fth fish, Pj_h(Wz_m f ) represents the position vector Wz_m of the fth fish f The value of the evaluation function at

[0088] Set the center of mass position of the fish school to be updated Among them, if the value of the evaluation function at the current centroid position is higher than the previous best position, the best position is updated, the number of iterations is set to U_KB, and it stops when the maximum number of iterations is reached. The current best position is extracted, and the exhibition form at this best position is set as the best exhibition form.

[0089] Using the best exhibition form to display exhibits includes: arranging the best exhibition form on the display terminal and monitoring the operating status of the exhibition form in real time.

[0090] Embodiment 2

[0091] Reference Figure 1 and Figure 2 , Embodiment 2 further illustrates a dynamic exhibition system for digital museum exhibits proposed by the present invention.

[0092] The method for dynamic exhibition of digital museum exhibits comprises the following steps:

[0093] Step S1: Collecting comprehensive original information, identifying text in the comprehensive original information, preprocessing the text and performing feature extraction to obtain a feature set; constructing a three-dimensional simulation model based on the appearance information of the exhibit; synapsing the feature set with the three-dimensional simulation model to form a comprehensive data set;

[0094] Step S2: construct and train an exhibit enrichment model based on Text2Shape, design a comprehensive data set of current exhibits to generate preset data, import the preset data into the trained exhibit enrichment model, and predict a multi-form set of exhibits;

[0095] Step S3: designing an optimization algorithm based on Fish School Search, identifying the best exhibition form from the multiple forms of exhibits through the optimization algorithm, and displaying the exhibits using the best exhibition form;

[0096] Step S4: Arrange the best exhibition form on the display terminal and monitor the operation status of the exhibition form in real time.

[0097] The dynamic exhibition method of digital museum exhibits, applied to a dynamic exhibition system of digital museum exhibits, comprises:

[0098] Information collection module: used to collect and process comprehensive original information to obtain a comprehensive data set;

[0099] Exhibition form generation module: build and train an exhibit enrichment model, input the comprehensive data set to generate preset data, import the preset data into the trained exhibit enrichment model, and obtain a multi-form set of exhibits;

[0100] Optimization algorithm module: Use optimization algorithm to identify the best exhibition form among multiple forms of exhibits, and use the best exhibition form to display the exhibits;

[0101] The modules are connected to each other via wired and / or wireless means.

[0102] In addition, according to the implementation of the present application, the process described in the figure of a dynamic exhibition system for digital museum exhibits can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided by the present application. Of course, the architecture shown in the figure of a dynamic exhibition system for digital museum exhibits is only exemplary. When implementing different devices, adaptive selection or adjustment can be made according to actual needs.

[0103] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0104] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A dynamic exhibition system for digital museum exhibits, characterized in that: The dynamic exhibition system of the digital museum exhibits includes: Information collection module: used to collect and process comprehensive original information to obtain a comprehensive data set; Exhibition form generation module: build and train an exhibit enrichment model, input the comprehensive data set to generate preset data, import the preset data into the trained exhibit enrichment model, and obtain a multi-form set of exhibits; Optimization algorithm module: Use optimization algorithm to identify the best exhibition form among multiple forms of exhibits, and use the best exhibition form to display the exhibits; The modules are connected to each other via wired and / or wireless means.

2. A dynamic exhibition system for digital museum exhibits according to claim 1, characterized in that: The comprehensive original information includes basic information of exhibits, evaluation feedback information and preview information. The data categories of basic information of exhibits include background information, appearance information and multimedia information of exhibits. The data categories of evaluation feedback information include audience behavior data and expert value assessment data. The data categories of preview information include the audience and display environment of exhibits.

3. A dynamic exhibition system for digital museum exhibits according to claim 2, characterized in that: The processing of the comprehensive original information to obtain the comprehensive data set specifically includes: Identify the text in the comprehensive original information, pre-process the text and perform feature extraction to obtain a feature set; Extract basic information of exhibits from comprehensive original information, and build a 3D simulation model based on the appearance information of exhibits; The feature set is synaptically coupled with the three-dimensional simulation model to form a comprehensive data set.

4. A dynamic exhibition system for digital museum exhibits according to claim 3, characterized in that: The method of obtaining a feature set after preprocessing the text and extracting features includes: Assume that the text of each data category includes R_t documents. Assume that the vocabulary C_h = {w1,…,wj,…,wn} is the set of all words in the entire text. Use TF-IDF to convert any document di into a TF-IDF value represented by a word frequency vector, and then generate a feature vector V for each document. di =[C_p(w1,di),……,C_p(wn,di)], where the TF-IDF value of word wj in document di is calculated TF(wj,di) represents the frequency of word wj in document di, DF(wj) is the number of documents in which word wj appears, that is, the document frequency, and i=1, ..., R_t, j=1, ..., n, Among them, r_tij represents the number of occurrences of word wj in document di, and w_dj represents the total number of all words in document di; Construct a P_Lasso regression model and set the objective function as Among them, β u_o is the regression coefficient of the u_oth TF-IDF value, γ is the regularization parameter used to control the sparsity of the model, and x w_o，u_o is the u_oth TF-IDF value of the w_oth word, W_F is the number of words, U_F is the number of TF-IDF values, and the component value of the w_oth word Among them, P(wj, Z_T) represents the joint probability of word wj and data category Z_Tb, that is, the probability of co-appearing in the document, P(wj) represents the marginal probability of word wj, that is, the probability of word wj appearing in the document, and P(Z_Tb) represents the marginal probability of the bth data category Z_Tb in the document; Using the feature vector as input and the component value of the word as output, the value of the regression coefficient is adjusted by training the P_Lasso regression model, and the words with non-zero regression coefficients are retained to obtain the feature set.

5. A dynamic exhibition system for digital museum exhibits according to claim 4, characterized in that: The method for preprocessing text comprises: Remove noise and stop words from the text, convert it into a unified format, and perform word segmentation after stem extraction and lemma restoration to obtain comprehensive original information after preprocessing.

6. A dynamic exhibition system for digital museum exhibits according to claim 5, characterized in that: The method for synapsing the feature set with the three-dimensional simulation model comprises: Identify each feature category data package in the feature set, match the feature category data package with the corresponding attributes of the 3D simulation model, set a click point at the corresponding location of the 3D simulation model match, hyperlink the click point to the corresponding feature category data package, add the feature category data package to the 3D simulation model of the exhibit, and form a comprehensive data set.

7. A dynamic exhibition system for digital museum exhibits according to claim 6, characterized in that: The method of building and training the exhibit enrichment model includes: A preset sample set, which includes o_p groups of data, each of which includes preset data of an exhibit and i_p exhibition forms corresponding to the exhibit; The exhibit enrichment model is designed based on Text2Shape. The conditional cVAE architecture is used to design the generator. The input of the generator includes preset data, and the output is the exhibition form generated based on the input. The loss function is set to in, For all noise vectors P Z (Z), D(G(Z, F_text, x_3d), F_text) represents the authenticity score of the discriminator on the three-dimensional shape G(Z, F_text, x_3d) and the text description F_text in the generated exhibition form, x_3d represents the three-dimensional shape, and Z represents the noise vector; The discriminator is designed to determine whether the generated exhibition form conforms to the input preset data. The input of the discriminator is the generated exhibition form. The loss function is set to in, Represents all real three-dimensional shapes P data The expected value of the true 3D shape x_3d in (x_3d), Represents all generated 3D shapes Producing 3D shapes The expected value of , D(x_3d, F_text) represents the authenticity score of the discriminator for the true 3D shape x_3d and the text description F_text, Represents the discriminator to generate a 3D shape and authenticity score of the text description F_text; The learning rate is set, and the sample set is divided into multiple batches and input into the exhibit enrichment model in sequence until the expected effect is achieved and the training is stopped to obtain a trained exhibit enrichment model.

8. A dynamic exhibition system for digital museum exhibits according to claim 7, characterized in that: The method of inputting the comprehensive data set into a design to generate preset data includes: For the text in the comprehensive dataset, a pre-trained language model is used to convert the text description into an embedding vector, which represents the semantic information of the text. Use a 3D convolutional neural network or a point cloud-based neural network to encode the 3D simulation model, convert the appearance information of the 3D simulation model into a dimensional vector, and extract the key features of the model; Collect the embedding vectors and dimension vectors of the exhibits to form preset data.

9. A dynamic exhibition system for digital museum exhibits according to claim 8, characterized in that: The method for identifying the best exhibition form in a set of multiple exhibit forms by using an optimization algorithm includes: The optimization algorithm is designed based on Fish School Search, and the optimization goal is defined as maximizing the value of the evaluation function. The number of fish schools A_z and the position vector Wz_m are initialized, and the evaluation function is Pj_h=∈1×∑ zj_u∈pd zj_u+∈2×∑ gz_k∈xw gz_k+∈3×hj_qh+∈4×qh_d, where zj_u represents the evaluation value of the u-th data category in the expert's value evaluation data pd, gz_k represents the evaluation value of the k-th data category in the audience's behavior data xw, hj_qh represents the fit between the exhibition form and the exhibition environment, qh_d represents the degree of innovation of the exhibition form, ∈1, ∈2, ∈3, ∈4 represent weight coefficients respectively; Set the foraging behavior: in, represents the position vector of the fth fish in the gth round, represents the position vector of the fth fish at the g-1th round, σ is the learning rate, if the value of the evaluation function of the fth fish at the g-1th round Less than the value of the evaluation function at round g Update for The movement direction of the fth fish Among them, Wz_m f represents the position vector of the fth fish, Wz_m l represents the feature vector of the lth fish in the neighborhood Nb(f) of the fth fish, Pj_h(Wz_m f ) represents the position vector Wz_m of the fth fish f The value of the evaluation function at Set the center of mass position of the fish school to be updated Among them, if the value of the evaluation function at the current centroid position is higher than the previous best position, the best position is updated, the number of iterations is set to U_KB, and it stops when the maximum number of iterations is reached. The current best position is extracted, and the exhibition form at this best position is set as the best exhibition form.

10. A dynamic exhibition system for digital museum exhibits according to claim 9, characterized in that: The method of displaying the exhibits in the best display form includes: arranging the best display form on the display terminal and monitoring the operation status of the display form in real time.

Images