Historical scenic spot-oriented generation type smart text travel guide system and historical scenic spot-oriented generation type smart text travel guide method

Abstract

The invention, which relates to the technical field of smart text travel, discloses a historical scenic spot-oriented generative smart text travel guide system comprising a historical scenic spot image acquisition module, a pre-training large model module, an image-text understanding module, a poem generation module and an output display module. The invention further discloses a historical scenic spot-oriented generation type smart text travel guide method. The method comprises the following steps of: acquiring an image; extracting feature information; constructing a large-model language knowledge structure, and understanding the cultural connotation of the historical scenic spot; comprehensive information about historical scenic spots is acquired; generating poems conforming to the cultural background and the poem rhythm of the historical scenic area; and displaying the generated poems to tourists through the AR technology. According to the invention, by integrating image acquisition and processing, a large language model, multi-modal data fusion and augmented reality technologies, immersive and personalized guide experience rich in historical culture connotation can be provided for tourists in historical scenic spots, and the visiting experience of the tourists is improved.

Description

Technical Field

[0001] The present invention relates to the field of smart cultural tourism technology, and in particular to a generative smart cultural tourism guide system and method for historical scenic spots. Background Art

[0002] With the rapid development of information technology, especially in the application of artificial intelligence and augmented reality technology, traditional tourism methods are undergoing profound changes. As one of the core of this change, smart cultural tourism is providing tourists with a more personalized and interactive tourism experience. Especially in historical scenic spots, how to enhance tourists' cultural experience and interactive perception through innovative technologies has become an important direction for the development of the cultural tourism industry. As an important part of cultural tourism, historical scenic spots usually have rich historical and cultural connotations and unique landscape resources. However, due to differences in cultural backgrounds and insufficient knowledge of tourists, tourists often find it difficult to fully understand the historical value, cultural background and stories behind the scenic spots, which affects the depth and quality of their tourism experience. In order to make up for this deficiency, the existing guide system mainly relies on manual explanation, QR code guide, voice guide and other methods, but the interactivity and immersion of these traditional methods are still limited, and it is difficult to fully mobilize the interest and emotional resonance of tourists.

[0003] However, the existing intelligent guide systems mainly focus on the single processing of text or images, lack the comprehensive use of multimodal information, and lack personalized content generation and deep emotional resonance experience based on historical and cultural backgrounds. For example, the creation of poems for historical scenic spots, the setting of emotional tones, and the matching with the cultural atmosphere of the scenic spots, the existing technical means are still not mature and accurate enough. Summary of the invention

[0004] In order to solve the above technical problems, a generative smart cultural and tourism guide system and method for historical scenic spots is provided. This technical solution solves the above problems.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is: A generative intelligent cultural tourism guide system and method for historical scenic spots, comprising: Historical scenic spot image acquisition module: used to collect photos and videos of historical scenic spots and extract image features; Pre-trained large model module: used for pre-training based on a large amount of text data to build language knowledge structures related to historical scenic spots; Graphics and text comprehension module: used to analyze and understand the photos and videos of historical scenic spots, and associate them with the historical and cultural background and relevant poetry knowledge to obtain a comprehensive understanding of historical scenic spots; Poetry generation module: Based on the output of the image-text understanding module and the language knowledge of the large model, it generates poems that conform to the cultural background of the historical scenic spot and determines the theme, emotional tone and rhythm requirements of the poem; Output display module: The generated poems are displayed to tourists in digital form through AR technology, and the poetry can be read aloud using speech synthesis technology.

[0006] Preferably, the historical scenic spot image acquisition module specifically includes: Image and video acquisition unit: Use cameras and drones to take photos and videos of historical scenic spots, capture the characteristics of scenic spots from different angles and time periods, and record the real-time dynamic situation of historical scenic spots, including tourist activities, seasonal changes, and scenic environment; Image preprocessing unit: remove blur and low-quality images, remove noise in images based on Gaussian filtering technology, optimize image resolution and contrast, remove and segment backgrounds for images containing interference elements, highlight the core elements of the scenic area, adjust image size, crop to the area of ​​interest, and divide the video into single-frame images. The Gaussian filtering formula is:

[0007] In the formula, is the standard deviation, is the spatial coordinate of the filter, is the value of the two-dimensional Gaussian function, is the circumference of a circle, is an exponential function; Image feature extraction unit: Identify specific objects in images based on convolutional neural network technology, including historical buildings, cultural relics, sculptures and garden landscapes; Architectural style recognition: Identify and classify ancient buildings of different styles and identify the types of buildings; Character and costume recognition: Detect and identify characters’ costumes, postures, and activities to help understand historical scenes and contexts; Landscape layout analysis: extract the overall layout of the scenic area, including gardens, mountains, lakes and roads, and analyze their structure; Natural element extraction: including the identification and extraction of natural landscapes such as mountains, rivers, and trees, to help depict the natural features of the scenic area; Color and light analysis: Analyze the color matching, light intensity and shadow distribution in the image to further enrich the understanding of the scenic environment.

[0008] Preferably, the pre-trained large model module specifically includes: Language model architecture selection and construction unit: Establish the model architecture for text understanding and generation tasks. Targeting historical scenic spot knowledge, select the GPT-3 model to process large-scale text data and capture semantic information in the context. Model initialization unit: Use the existing pre-trained model as the initial weight, initialize it to a multi-task learning architecture, train multiple related tasks at the same time, and define the pre-training objectives; Historical background learning task unit: Design specific tasks based on the background of historical scenic spots, use knowledge-enhanced language modeling, and embed knowledge graphs and entity information related to historical scenic spots into the model; Multimodal data fusion and multi-task learning unit: language data is integrated with other modal data for multimodal learning, which improves the model's ability to understand historical scenic spots. At the same time, the model is trained to perform multiple tasks, including historical event reasoning, travel recommendations, and historical figure relationship reasoning. The data fusion formula is:

[0009] In the formula, For the final fusion result, is the number of modes, is the modal index, is the modal weight, It is The feature representation of the modality.

[0010] Preferably, the graphic and text understanding module specifically includes: Historical context association unit: Identify the specific location of historical scenic spots based on clues in the image, and match the scenery in the image and video with the known historical and cultural background; Poetry and Literature Association Unit: Based on natural language processing technology, it finds verses and descriptions related to scenic spots from poems and ancient Chinese literary works, automatically infers historical and cultural backgrounds based on image content, and matches poems, idioms and famous sayings related to the scenic spot; Sentiment Analysis and Context Extraction Unit: Through sentiment analysis of poems, historical documents and images, we can further understand the emotional expression of these cultural works on historical scenic spots; Multimodal knowledge fusion unit: integrates image content with poetry and historical background knowledge to generate comprehensive information; Semantic understanding and reasoning unit: By constructing a knowledge graph related to the scenic spot, we can achieve an in-depth understanding of the cultural background of the scenic spot, and combine image and text information for reasoning and analysis.

[0011] Preferably, the poetry generation module specifically includes: Image and text understanding module output parsing unit: Based on image analysis, it extracts the visual elements of the scenic spot, the emotional atmosphere and historical and cultural information of the scenic spot, and provides emotional and contextual background for poetry creation; Theme determination and emotional tone setting unit: Determine the theme of the poem based on the historical background, cultural characteristics and graphic output of the scenic spot, and choose the emotional tone of the poem based on the emotional atmosphere of the scenic spot; Meter and style requirements setting unit: choose the appropriate meter of poetry according to the theme and emotional tone of the poem; Language model generation unit: Generate poems that are consistent with the cultural background of historical scenic spots based on the large language model. The theme, emotional tone, rhythm requirements, and the output of the picture and text understanding module are passed as input conditions to the large language model, and the large language model is used for poetry generation.

[0012] Preferably, the output display module specifically includes: Augmented reality technology integration unit: Use AR equipment to perceive the scenic area environment in real time, determine the location of the poem display in combination with the location information of the scenic area, use the buildings, natural landscapes and historical sites of the scenic area as the background, and use the AR engine to superimpose the poem content on the actual scene. Superimpose the generated poem on the real scene of the scenic area in the form of a virtual layer. Visitors interact with the AR displayed poem through gestures and voice; Speech synthesis technology integration unit: selects the speech synthesis engine, configures the language, intonation, speed and emotional tone of the speech, divides the generated poetry text into speech paragraphs, adjusts the tone, intonation and speed parameters of speech synthesis according to the emotional tone of the poem, and synchronizes the speech reading with the display content of the poem.

[0013] A generative intelligent cultural tourism guiding method for historical scenic spots, comprising: S1: Image acquisition, obtaining photos and videos of historical scenic spots through image acquisition equipment and transmitting them to the system; S2: Image feature extraction, using image recognition technology to extract architectural style and landscape layout feature information from historical scenic spot photos; S3: Large model pre-training, through learning massive historical and cultural text data, building a large model of language knowledge structure, to understand the cultural connotation of historical scenic spots; S4: Image and text understanding and knowledge association, associate and understand image features with historical and cultural knowledge in the big model to obtain comprehensive information about historical scenic spots; S5: Poetry generation: Based on the information provided by the graphic comprehension module, the theme and emotional tone of the poem are determined, and poems that conform to the cultural background of the historical scenic spot and the rhythm of the poem are generated; S6: Display and interaction: The generated poems are presented to tourists through AR technology and read aloud through speech synthesis technology to enhance tourists’ sense of immersion and optimize the guided tour experience based on tourists’ feedback.

[0014] The specific steps of S1 are: Use cameras and drones to take photos and videos of historical scenic spots to capture the characteristics of the scenic spots and record real-time dynamic conditions; Set shooting time and period to capture different light and seasonal changes, formulate shooting routes and angles according to the characteristics of historical scenic spots, and ensure that all important scenic spots, buildings and gardens are fully covered; Real-time filming of scenic spots, including dynamic elements of visitor activities, seasonal changes, and weather conditions, and capturing high-definition photos and videos of static elements of historical buildings, sculptures, and garden landscapes; Organize and store the captured data to ensure data integrity and security; The image quality detection algorithm is used to automatically filter out blurry and poor quality images for elimination, and Gaussian filtering is used to denoise the images; The image resolution and contrast are adjusted through image enhancement technology to make the image clearer and the details more prominent. The background of the image containing interference elements is removed and the image is segmented to highlight the core elements of the scenic spot. Adjust the image size and cut the video file into single-frame images; Through convolutional neural network technology, key features in images, including historical buildings, people and natural elements, can be automatically identified and extracted.

[0015] Preferably, the specific steps of S2 are: Choose GPT-3 as the infrastructure, generate text through contextual inference, process complex texts of historical scenic spots, and determine the tasks that the model needs to perform; Provide the model with relevant corpus of historical scenic spots, including information about historical buildings, events, and people; Use GPT-3’s pre-trained weights to initialize the model, fine-tune its weights based on task requirements, and integrate multiple tasks into a unified multi-task learning framework; Set pre-training goals for each task and allocate computing resources based on the complexity and requirements of the task; Design historical background learning tasks, embed knowledge graphs and entity information related to historical scenic spots into the model, and use graph embedding technology to convert structured data into a form that the model can process; The language data is integrated with other modal data to conduct multimodal learning, and the model's comprehensive understanding of historical scenic spots is improved through multi-task learning, integrating the features of different modalities.

[0016] Preferably, the specific steps of S3 are: Identify the specific location of historical sites based on clues in images, and match the scenes in images and videos with known historical and cultural contexts; Based on natural language processing technology, we can find poems and descriptions related to scenic spots from poems and ancient Chinese literature, automatically infer the historical and cultural background based on the image content, and match poems, idioms and famous sayings related to scenic spots. Through sentiment analysis of poems, historical documents and images, we can further understand the emotional expressions of these cultural works on historical sites; Integrate image content with poetry and historical background knowledge to generate comprehensive information; By constructing a knowledge graph related to the scenic spot, we can achieve an in-depth understanding of the cultural background of the scenic spot, and combine image and text information for reasoning and analysis.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention proposes to provide an immersive, personalized and historically and culturally rich guided tour experience for tourists in historical scenic spots by integrating image acquisition and processing, large language models, multimodal data fusion and augmented reality technology. This system can not only enhance the tourists' visiting experience, but also promote the digital and intelligent application of cultural heritage, and has important technical and commercial value. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a system framework diagram of the present invention; Figure 2 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0019] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0020] Reference Figure 1 As shown, a generative intelligent cultural tourism guide system and method for historical scenic spots includes: Historical scenic spot image acquisition module: Image and video acquisition unit: Use cameras and drones to take photos and videos of historical scenic spots, capture the characteristics of scenic spots from different angles and time periods, and record the real-time dynamic situation of historical scenic spots, including tourist activities, seasonal changes, and scenic environment; Image preprocessing unit: remove blur and low-quality images, remove noise in images based on Gaussian filtering technology, optimize image resolution and contrast, remove and segment backgrounds for images containing interference elements, highlight the core elements of the scenic area, adjust image size, crop to the area of ​​interest, and divide the video into single-frame images. The Gaussian filtering formula is:

[0021] In the formula, is the standard deviation, is the spatial coordinate of the filter, is the value of the two-dimensional Gaussian function, is the circumference of a circle, is an exponential function; Image feature extraction unit: Identify specific objects in images based on convolutional neural network technology, including historical buildings, cultural relics, sculptures and garden landscapes; Architectural style recognition: Identify and classify ancient buildings of different styles and identify the types of buildings; Character and costume recognition: Detect and identify characters’ costumes, postures, and activities to help understand historical scenes and contexts; Landscape layout analysis: extract the overall layout of the scenic area, including gardens, mountains, lakes and roads, and analyze their structure; Natural element extraction: including the identification and extraction of natural landscapes such as mountains, rivers, and trees, to help depict the natural features of the scenic area; Color and light analysis: Analyze the color matching, light intensity and shadow distribution in the image to further enrich the understanding of the scenic environment.

[0022] Pre-trained large model module: Preferably, the pre-trained large model module specifically includes: Language model architecture selection and construction unit: Establish the model architecture for text understanding and generation tasks. Targeting historical scenic spot knowledge, select the GPT-3 model to process large-scale text data and capture semantic information in the context. Model initialization unit: Use the existing pre-trained model as the initial weight, initialize it to a multi-task learning architecture, train multiple related tasks at the same time, and define the pre-training objectives; Historical background learning task unit: Design specific tasks based on the background of historical scenic spots, use knowledge-enhanced language modeling, and embed knowledge graphs and entity information related to historical scenic spots into the model; Multimodal data fusion and multi-task learning unit: language data is integrated with other modal data for multimodal learning, which improves the model's ability to understand historical scenic spots. At the same time, the model is trained to perform multiple tasks, including historical event reasoning, travel recommendations, and historical figure relationship reasoning. The data fusion formula is:

[0023] In the formula, For the final fusion result, is the number of modes, is the modal index, is the modal weight, It is The feature representation of the modality.

[0024] Graphical comprehension module: Historical context association unit: Identify the specific location of historical scenic spots based on clues in the image, and match the scenery in the image and video with the known historical and cultural background; Poetry and Literature Association Unit: Based on natural language processing technology, it finds verses and descriptions related to scenic spots from poems and ancient Chinese literary works, automatically infers historical and cultural backgrounds based on image content, and matches poems, idioms and famous sayings related to the scenic spot; Sentiment Analysis and Context Extraction Unit: Through sentiment analysis of poems, historical documents and images, we can further understand the emotional expression of these cultural works on historical scenic spots; Multimodal knowledge fusion unit: integrates image content with poetry and historical background knowledge to generate comprehensive information; Semantic understanding and reasoning unit: By constructing a knowledge graph related to the scenic spot, we can achieve an in-depth understanding of the cultural background of the scenic spot, and combine image and text information for reasoning and analysis.

[0025] Poetry generation module: Image and text understanding module output parsing unit: Based on image analysis, it extracts the visual elements of the scenic spot, the emotional atmosphere and historical and cultural information of the scenic spot, and provides emotional and contextual background for poetry creation; Theme determination and emotional tone setting unit: Determine the theme of the poem based on the historical background, cultural characteristics and graphic output of the scenic spot, and choose the emotional tone of the poem based on the emotional atmosphere of the scenic spot; Meter and style requirements setting unit: choose the appropriate meter of poetry according to the theme and emotional tone of the poem; Language model generation unit: Generate poems that are consistent with the cultural background of historical scenic spots based on the large language model. The theme, emotional tone, rhythm requirements, and the output of the picture and text understanding module are passed as input conditions to the large language model, and the large language model is used for poetry generation.

[0026] Output display module: Augmented reality technology integration unit: Use AR equipment to perceive the scenic area environment in real time, determine the location of the poem display in combination with the location information of the scenic area, use the buildings, natural landscapes and historical sites of the scenic area as the background, and use the AR engine to superimpose the poem content on the actual scene. Superimpose the generated poem on the real scene of the scenic area in the form of a virtual layer. Visitors interact with the AR displayed poem through gestures and voice; Speech synthesis technology integration unit: selects the speech synthesis engine, configures the language, intonation, speed and emotional tone of the speech, divides the generated poetry text into speech paragraphs, adjusts the tone, intonation and speed parameters of speech synthesis according to the emotional tone of the poem, and synchronizes the speech reading with the display content of the poem.

[0027] A generative intelligent cultural tourism guiding method for historical scenic spots, comprising: S1: Use cameras and drones to take photos and videos of historical scenic spots to capture the characteristics of the scenic spots and record real-time dynamic conditions; Set shooting time and period to capture different light and seasonal changes, formulate shooting routes and angles according to the characteristics of historical scenic spots, and ensure that all important scenic spots, buildings and gardens are fully covered; Real-time filming of scenic spots, including dynamic elements of visitor activities, seasonal changes, and weather conditions, and capturing high-definition photos and videos of static elements of historical buildings, sculptures, and garden landscapes; Organize and store the captured data to ensure data integrity and security; The image quality detection algorithm is used to automatically filter out blurry and poor quality images for elimination, and Gaussian filtering is used to denoise the images; The image resolution and contrast are adjusted through image enhancement technology to make the image clearer and the details more prominent. The background of the image containing interference elements is removed and the image is segmented to highlight the core elements of the scenic spot. Adjust the image size and cut the video file into single-frame images; Through convolutional neural network technology, key features in images, including historical buildings, people and natural elements, can be automatically identified and extracted.

[0028] S2: Choose GPT-3 as the infrastructure, generate text through contextual inference, process complex texts of historical scenic spots, and determine the tasks that the model needs to perform; Provide the model with relevant corpus of historical scenic spots, including information about historical buildings, events, and people; Use GPT-3’s pre-trained weights to initialize the model, fine-tune its weights based on task requirements, and integrate multiple tasks into a unified multi-task learning framework; Set pre-training goals for each task and allocate computing resources based on the complexity and requirements of the task; Design historical background learning tasks, embed knowledge graphs and entity information related to historical scenic spots into the model, and use graph embedding technology to convert structured data into a form that the model can process; The language data is integrated with other modal data to conduct multimodal learning, and the model's comprehensive understanding of historical scenic spots is improved through multi-task learning, integrating the features of different modalities.

[0029] S3: Identify the specific location of historical sites based on clues in images, and match the scenes in images and videos with known historical and cultural contexts; Based on natural language processing technology, we can find poems and descriptions related to scenic spots from poems and ancient Chinese literature, automatically infer the historical and cultural background based on the image content, and match poems, idioms and famous sayings related to scenic spots. Through sentiment analysis of poems, historical documents and images, we can further understand the emotional expressions of these cultural works on historical sites; Integrate image content with poetry and historical background knowledge to generate comprehensive information; By constructing a knowledge graph related to the scenic spot, we can achieve an in-depth understanding of the cultural background of the scenic spot, and combine image and text information for reasoning and analysis.

[0030] S4: Graphics and text understanding and knowledge association, associate and understand image features with historical and cultural knowledge in the big model to obtain comprehensive information about historical scenic spots.

[0031] S5: Poetry generation, based on the information provided by the graphic comprehension module, determines the theme and emotional tone of the poem, and generates poems that conform to the cultural background of the historical scenic spot and the rhythm of the poem.

[0032] S6: Display and interaction: The generated poems are presented to tourists through AR technology and read aloud through speech synthesis technology to enhance tourists' sense of immersion and optimize the guided tour experience based on tourists' feedback.

[0033] In summary, the advantages of the present invention are: By combining the visual elements of historical scenic spots with cultural background knowledge, generating poetry content that fits the historical atmosphere, it can enable tourists to have a deeper understanding of the history and culture of the scenic spots and feel the charm of culture; The system can automatically generate historical and cultural poems related to different tourists' interests based on their interests, preferences and feedback, and provide personalized guided tour services through AR and voice technology. Tourists can interact with the displayed poems, which enhances the personalization and fun of the guided tour. Through automated historical background reasoning and cultural association, visitors can receive in-depth historical education during their visit. The system not only provides a visual tour, but also enhances visitors’ historical and cultural knowledge through rich cultural content such as poetry, idioms, historical figures and events. By automatically generating historical and cultural background and poetry, the system can not only reduce the workload of traditional tour guides, but also provide efficient and timely guided tour services in scenic spots with high tourist flow, thus improving the tourist experience; The present invention combines modern artificial intelligence technology and augmented reality technology, creating a new tour guide mode in the field of cultural tourism, which is highly innovative and technologically advanced; The system can be adjusted according to different historical scenic spots, cultural themes and tourists' needs, and has high scalability. Whether it is an ancient building, natural landscape or cultural site, it can generate poetry and information that conforms to its historical and cultural background through image and text analysis. It is suitable for cultural guides in a variety of historical scenic spots.

[0034] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A generative intelligent cultural tourism guide system for historical scenic spots, characterized in that: include: Historical scenic spot image acquisition module: used to collect photos and videos of historical scenic spots and extract image features; Pre-trained large model module: used for pre-training based on a large amount of text data to build language knowledge structures related to historical scenic spots; Graphics and text comprehension module: used to analyze and understand the photos and videos of historical scenic spots, and associate them with the historical and cultural background and relevant poetry knowledge to obtain a comprehensive understanding of historical scenic spots; Poetry generation module: Based on the output of the image-text understanding module and the language knowledge of the large model, it generates poems that conform to the cultural background of the historical scenic spot and determines the theme, emotional tone and rhythm requirements of the poem; Output display module: The generated poems are displayed to tourists in digital form through AR technology, and the poetry can be read aloud using speech synthesis technology.

2. A generative intelligent cultural tourism guide system and method for historical scenic spots according to claim 1, characterized in that: The historical scenic spot image acquisition module specifically includes: Image and video acquisition unit: Use cameras and drones to take photos and videos of historical scenic spots, capture the characteristics of scenic spots from different angles and time periods, and record the real-time dynamic situation of historical scenic spots, including tourist activities, seasonal changes, and scenic environment; Image preprocessing unit: remove blur and low-quality images, remove noise in images based on Gaussian filtering technology, optimize image resolution and contrast, remove and segment backgrounds for images containing interference elements, highlight the core elements of the scenic area, adjust image size, crop to the area of ​​interest, and divide the video into single-frame images. The Gaussian filtering formula is: In the formula, is the standard deviation, is the spatial coordinate of the filter, is the value of the two-dimensional Gaussian function, is the circumference of a circle, is an exponential function; Image feature extraction unit: Based on convolutional neural network technology, it can identify specific objects in images, including historical buildings, cultural relics, sculptures and garden landscapes.

3. According to claim 2, a generative intelligent cultural tourism guide system for historical scenic spots is characterized in that: The pre-trained large model module specifically includes: Language model architecture selection and construction unit: Establish the model architecture for text understanding and generation tasks. Targeting historical scenic spot knowledge, select the GPT-3 model to process large-scale text data and capture semantic information in the context. Model initialization unit: Use the existing pre-trained model as the initial weight, initialize it to a multi-task learning architecture, train multiple related tasks at the same time, and define the pre-training objectives; Historical background learning task unit: Design specific tasks based on the background of historical scenic spots, use knowledge-enhanced language modeling, and embed knowledge graphs and entity information related to historical scenic spots into the model; Multimodal data fusion and multi-task learning unit: language data is integrated with other modal data for multimodal learning, which improves the model's ability to understand historical scenic spots. At the same time, the model is trained to perform multiple tasks, including historical event reasoning, travel recommendations, and historical figure relationship reasoning. The data fusion formula is: In the formula, For the final fusion result, is the number of modes, is the modal index, is the modal weight, It is The feature representation of the modality.

4. A generative intelligent cultural tourism guide system for historical scenic spots according to claim 3, characterized in that: The graphic and text understanding module specifically includes: Historical context association unit: Identify the specific location of historical scenic spots based on clues in the image, and match the scenery in the image and video with the known historical and cultural background; Poetry and Literature Association Unit: Based on natural language processing technology, it finds verses and descriptions related to scenic spots from poems and ancient Chinese literary works, automatically infers historical and cultural backgrounds based on image content, and matches poems, idioms and famous sayings related to the scenic spot; Sentiment Analysis and Context Extraction Unit: Through sentiment analysis of poems, historical documents and images, we can further understand the emotional expression of these cultural works on historical scenic spots; Multimodal knowledge fusion unit: integrates image content with poetry and historical background knowledge to generate comprehensive information; Semantic understanding and reasoning unit: By constructing a knowledge graph related to the scenic spot, we can achieve an in-depth understanding of the cultural background of the scenic spot, and combine image and text information for reasoning and analysis.

5. According to claim 4, a generative intelligent cultural tourism guide system for historical scenic spots is characterized in that: The poetry generation module specifically includes: Image and text understanding module output parsing unit: Based on image analysis, it extracts the visual elements of the scenic spot, the emotional atmosphere and historical and cultural information of the scenic spot, and provides emotional and contextual background for poetry creation; Theme determination and emotional tone setting unit: Determine the theme of the poem based on the historical background, cultural characteristics and graphic output of the scenic spot, and choose the emotional tone of the poem based on the emotional atmosphere of the scenic spot; Meter and style requirements setting unit: choose the appropriate meter of poetry according to the theme and emotional tone of the poem; Language model generation unit: Generate poems that are consistent with the cultural background of historical scenic spots based on the large language model. The theme, emotional tone, rhythm requirements, and the output of the picture and text understanding module are passed as input conditions to the large language model, and the large language model is used for poetry generation.

6. A generative intelligent cultural tourism guide system for historical scenic spots according to claim 5, characterized in that: The output display module specifically includes: Augmented reality technology integration unit: Use AR equipment to perceive the scenic area environment in real time, determine the location of the poem display in combination with the location information of the scenic area, use the buildings, natural landscapes and historical sites of the scenic area as the background, and use the AR engine to superimpose the poem content on the actual scene. Superimpose the generated poem on the real scene of the scenic area in the form of a virtual layer. Visitors interact with the AR displayed poem through gestures and voice; Speech synthesis technology integration unit: selects the speech synthesis engine, configures the language, intonation, speed and emotional tone of the speech, divides the generated poetry text into speech paragraphs, adjusts the tone, intonation and speed parameters of speech synthesis according to the emotional tone of the poem, and synchronizes the speech reading with the display content of the poem.

7. A generative intelligent cultural tourism guide method for historical scenic spots, characterized in that: include: S1: Image acquisition, obtaining photos and videos of historical scenic spots through image acquisition equipment and transmitting them to the system; S2: Image feature extraction, using image recognition technology to extract architectural style and landscape layout feature information from historical scenic spot photos; S3: Large model pre-training, through learning massive historical and cultural text data, building a large model of language knowledge structure, to understand the cultural connotation of historical scenic spots; S4: Image and text understanding and knowledge association, associate and understand image features with historical and cultural knowledge in the big model to obtain comprehensive information about historical scenic spots; S5: Poetry generation: Based on the information provided by the graphic comprehension module, the theme and emotional tone of the poem are determined, and poems that conform to the cultural background of the historical scenic spot and the rhythm of the poem are generated; S6: Display and interaction: The generated poems are presented to tourists through AR technology and read aloud through speech synthesis technology to enhance tourists’ sense of immersion and optimize the guided tour experience based on tourists’ feedback.

8. A generative intelligent cultural and tourism guiding method for historical scenic spots according to claim 7, characterized in that: The specific steps of S1 are: Use cameras and drones to take photos and videos of historical scenic spots to capture the characteristics of the scenic spots and record real-time dynamic conditions; Set shooting time and period to capture different light and seasonal changes, formulate shooting routes and angles according to the characteristics of historical scenic spots, and ensure that all important scenic spots, buildings and gardens are fully covered; Real-time filming of scenic spots, including dynamic elements of visitor activities, seasonal changes, and weather conditions, and capturing high-definition photos and videos of static elements of historical buildings, sculptures, and garden landscapes; Organize and store the captured data to ensure data integrity and security; The image quality detection algorithm is used to automatically filter out blurry and poor quality images for elimination, and Gaussian filtering is used to denoise the images; The image resolution and contrast are adjusted through image enhancement technology to make the image clearer and the details more prominent. The background of the image containing interference elements is removed and the image is segmented to highlight the core elements of the scenic spot. Adjust the image size and cut the video file into single-frame images; Through convolutional neural network technology, key features in images, including historical buildings, people and natural elements, can be automatically identified and extracted.

9. A generative intelligent cultural tourism guide method for historical scenic spots according to claim 8, characterized in that: The specific steps of S2 are: Choose GPT-3 as the infrastructure, generate text through contextual inference, process complex texts of historical scenic spots, and determine the tasks that the model needs to perform; Provide the model with relevant corpus of historical scenic spots, including information about historical buildings, events, and people; Use GPT-3’s pre-trained weights to initialize the model, fine-tune its weights based on task requirements, and integrate multiple tasks into a unified multi-task learning framework; Set pre-training goals for each task and allocate computing resources based on the complexity and requirements of the task; Design historical background learning tasks, embed knowledge graphs and entity information related to historical scenic spots into the model, and use graph embedding technology to convert structured data into a form that the model can process; The language data is integrated with other modal data to conduct multimodal learning, and the model's comprehensive understanding of historical scenic spots is improved through multi-task learning, integrating the features of different modalities.

10. A generative intelligent cultural tourism guide method for historical scenic spots according to claim 9, characterized in that: The specific steps of S3 are: Identify the specific location of historical sites based on clues in images, and match the scenes in images and videos with known historical and cultural contexts; Based on natural language processing technology, we can find poems and descriptions related to scenic spots from poems and ancient Chinese literature, automatically infer the historical and cultural background based on the image content, and match poems, idioms and famous sayings related to scenic spots. Through sentiment analysis of poems, historical documents and images, we can further understand the emotional expressions of these cultural works on historical sites; Integrate image content with poetry and historical background knowledge to generate comprehensive information; By constructing a knowledge graph related to the scenic spot, we can achieve an in-depth understanding of the cultural background of the scenic spot, and combine image and text information for reasoning and analysis.

Images