User travel willingness quantification method based on short video visual content presentation style

Abstract

The invention discloses a user travel willingness quantification method based on a short video visual content presentation style, and the method comprises the steps: collecting travel short videos, extracting key frames through an inter-frame difference method, carrying out the data marking, carrying out the model optimization through the powerful feature extraction capability of a pre-training model Inception-v3, and removing a top layer, thereby achieving the quantification of the travel willingness of a user. A global average pooling layer, a self-defined full-connection layer and an output layer are added, and image content is deeply analyzed to identify visual presentation styles of short videos, including portraits, landscapes and the like. According to the method, the potential influence of the visual style on the travel willingness of the user is analyzed, the travel willingness content of the user is detected and quantified, whether the user generates the travel willingness or not can be accurately recognized, the intention intensity is further evaluated, a content creator and a platform are helped to optimize video content, and the marketing effect and the user participation degree of travel-related content are improved.

Description

Technical Field

[0001] The present invention relates to the fields of image classification and short video analysis, and in particular to a method for quantifying user travel intention based on the presentation style of short video visual content. Background Art

[0002] With the advent of the digital age, the fields of cultural entertainment and tourism consumption have undergone major changes. In particular, short video platforms, such as TikTok, have become an important tool for promoting tourist destinations and building tourism brands. With their low cost and rapid dissemination, short videos have shown significant effects in attracting audiences and enhancing travel intentions.

[0003] In addition, combined with new media forms such as virtual reality (VR), short videos provide users with an immersive travel experience, thus changing the traditional tourism marketing model to some extent.

[0004] In the existing technology, although existing studies have confirmed the effectiveness of short videos in tourism marketing, most studies focus on analyzing text content and audience interaction, such as comments and online word of mouth, and often ignore the impact of visual content on travel intention. Although some studies focus on shooting perspectives and visual factors that trigger travel inspiration, these studies mainly rely on subjective audience feedback, so there may be certain deviations.

[0005] Therefore, further in-depth analysis of the impact of visual elements in short videos is needed to provide a more precise and objective analysis framework to quantify users' travel intentions and to objectively evaluate how different visual presentation styles actually affect viewers' travel intentions. Summary of the invention

[0006] The problem to be solved by the present invention is to provide a method for quantifying users' travel intention based on the presentation style of short video visual content, use a machine learning image classification model to analyze two key visual scenes, portraits and scenery, and objectively evaluate how the visual content in short videos affects the audience's travel intention.

[0007] The present invention adopts the following technical solution: a method for quantifying user travel intention based on the presentation style of short video visual content, comprising the following steps: Step S1, data collection and image preprocessing: collect tourism short video samples and perform image preprocessing to construct a sample data set; Step S2, key frame extraction and labeling: for the short video data in the sample data set, the local maximum is extracted by the frame difference method as the key frame of each short video, and some data is randomly selected from the key frame for labeling; Step S3, tourism intention coding: based on the comment information of short videos in the sample data set, randomly select some data, classify tourism intention through the three-level coding method, and count the number of tourism intention comments of all samples in the sample data set through the keyword matching method; Step S4, Inception-v3 model training and evaluation: Divide the preprocessed image data into a training set, a validation set, and a test set, adjust the pretrained Inception-v3 model, remove the top layer, add a global average pooling layer, a custom fully connected layer, and an output layer, use the validation set to evaluate the performance during training, adjust the Inception-v3 model parameters according to the performance chart, and verify the classification accuracy and generalization ability of the Inception-v3 model through the test set; Step S5, image classification and style judgment: input the image to be classified into the Inception-v3 model trained in step S4, automatically extract image features and classify them, and output a binary result for each classification, which respectively indicates that the image does not contain or contains a specific target element. Based on the classification result of the Inception-v3 model, the visual presentation style of the image is judged; Step S6, quantitative analysis of impact: Based on the image classification results obtained in step S5 and the number of travel intention comments in the video obtained in step S3, a multivariate linear regression model is constructed to quantitatively analyze the visual presentation style of the short video and its impact on the user's travel intention.

[0008] Preferably, in step S1, the travel short video samples include: comments, number of likes, number of favorites, number of shares, number of fans, video length and release time information; the image preprocessing includes: deleting duplicate data, correcting missing values, and removing text and image videos.

[0009] Preferably, in step S2, key frame extraction and labeling includes the following sub-steps: S2.1, split the short video data in the sample data set into frame sequences frame by frame, calculate the pixel difference between each frame and the previous frame, and identify the significant changes between frames; S2.2, using the moving average method to smooth the difference between frames, using the local maximum method to detect key change points in the smoothed difference data, and extracting the key frames in the video; S2.3. Randomly select part of the data from the key frames according to a preset ratio, perform manual annotation, form a standardized data set, load the key frame data in the standardized data set and perform image data processing, including: color correction, size adjustment, contrast enhancement, and set the image resolution to a uniform size, maintain the color format, and scale the image pixel value to between 0 and 1.

[0010] Preferably, the three-level coding method in step S3 includes: open coding, principal axis coding and selective coding; the tourism intention classification includes: expressing interest and plan, asking for information about tourist destinations, and interactive behavior.

[0011] Preferably, in step S3, the tourism intention encoding includes the following sub-steps: S3.1. Based on the comment information of short videos in the labeled sample data set, randomly select some data, screen out the comments that reflect travel intentions, and perform preliminary coding.

[0012] S3.2. Classify the content of travel intention comments by three-level coding method: S3.2.1. In the open coding stage, the original review information was read word by word using software to conceptualize the initial concepts. The double coding method was used to mark and code them respectively. After comparing and unifying the coding results, the initial concepts were classified. S3.2.2. In the axial coding stage, categories with similar and related meanings were logically classified and merged to form supercategories; S3.2.3. In the selective coding stage, the supercategories formed in the main axis coding stage were aggregated into core categories, including: strong intention, exploration interest, specific plan, action preparation, play activities, location inquiry, contact information, cost information, travel preparation, transportation route, finding companions and social activity participation; S3.2.4. Summarize the core of users’ travel intentions: Integrate the comments in the short video comment area and classify the short video comments containing travel intentions.

[0013] S3.3. Count the number of travel intention comments of all samples in the sample data set by keyword matching method: S3.3.1. Keyword selection based on coding categories: Based on the classification results of short video comments containing travel intentions in step S3.2.4, select several keywords representing the core content of each category; S3.3.2, Keyword matching: Keyword matching is performed in all the comments in the sample data set, and each comment containing one or more keywords is marked as a comment containing a specific travel intention; S3.3.3. Statistical analysis: Conduct statistical analysis on the number of comments containing travel intentions in a single short video, quantify the distribution of users’ interests in different tourist destinations, and the specific concerns and behavior patterns of users when expressing their travel intentions.

[0014] Preferably, in step S3.2.4, the short video comments containing travel intention are classified into: expressing interest and plan, asking for travel destination information, and interactive behavior.

[0015] Comments involving users expressing interest in a travel destination or future travel intentions were grouped into the expression of interest and plans category; The behavior of seeking information such as the location, cost, and guide of a specific tourist destination is classified as asking for tourist destination information; Behaviors that interact with others, such as finding companions and participating in social activities, are defined as interactive behavior.

[0016] Preferably, in step S4, adjusting the pre-trained Inception-v3 model includes the following sub-steps: S4.1. Based on the pre-trained Inception-v3 model, load the model without the top layer and set the base layer to be non-trainable. S4.2. Add a custom output layer, including a global average pooling layer and a fully connected layer, as well as a binary sigmoid activation output layer; S4.3. Compile using the Adam optimizer and binary cross entropy loss function, using early stopping to prevent overfitting. S4.4. Enhance the generalization ability of the Inception-v3 model through data augmentation and category weight adjustment.

[0017] Preferably, the data processing method of the adjusted Inception-v3 model is as follows: The feature map data stream is fed into the adjusted Inception-v3 model, starting with an input layer of size 299×299×3, suitable for RGB image data; The input data stream enters a series of base layers consisting of convolutional layers, pooling layers, and concatenation layers, which change the data dimension and gradually extract and combine feature maps of different scales.

[0018] The first base layer contains three convolutional pooling modules, each of which contains 1×1, 3×3, 5×5 convolution kernels and a maximum pooling layer. Each module first uses 1×1 convolution to reduce the number of channels, and then applies 3×3 and 5×5 convolutions to expand the receptive field respectively. At the same time, the maximum pooling layer is combined to capture spatial information at different levels, and the results are merged in the splicing layer; The second base layer, used to deepen the network, includes five convolutional pooling modules, each of which adds a 7×7 convolution kernel to capture a wider range of contextual information; The third basic layer refines feature extraction through several convolutional pooling modules.

[0019] After the input data flows through the base layer, it is passed to the global average pooling layer, which compresses the entire spatial dimension of each feature map into a single value. For each feature map, the average of all pixel values ​​is calculated and a fixed-size vector is output.

[0020] The fixed-size vector is fed into one or more fully connected layers to learn higher-level feature representations; finally, a binary classification task is performed through an output layer with a sigmoid activation function to produce the final prediction result.

[0021] Preferably, in step S5, the Inception-v3 model performs classification based on image features, and outputs a binary result of 0 or 1 for each classification, indicating that the image does not contain or contains the specific target element, respectively.

[0022] The visual presentation style includes: experience style and landscape style. If the target elements of the image are mainly close-ups of portraits, it is classified as experience style; if the target elements of the image are mainly landscape displays, it is classified as landscape style.

[0023] Preferably, in step S6, the multiple linear regression model is a generalized linear model Poisson regression, and the variables of the multiple linear regression model include: the number of comments on user travel intentions , the number of pictures with mainly landscape in the video keyframes , the number of images with portraits as the main feature in the video keyframes , video duration The time interval from video publishing to video collection , the number of followers of the publisher , video likes , the number of video comments , video sharing number , video collection number .

[0024] The relationship between variables is analyzed through a multiple regression model, and the degree of influence of variables on the number of comments on tourism intention is quantified through regression parameters β0~β8; by determining the setting principles of regression parameters and statistical significance, it is tested to determine which variables have a significant impact on the number of comments on tourism intention.

[0025] Specifically, we first determine the setting principle of regression parameters, that is, to determine which variables have a significant impact on the number of travel intention reviews through statistical significance tests. For example, if the regression coefficient of a variable is significantly not zero, it is considered that the variable has a significant impact on the number of travel intention reviews.

[0026] In this way, the following conclusions are drawn: S6.1. Number of landscape images in video key frames and number of portrait images It has a significant impact on the number of comments on travel intention; S6.2, Frequency duration The time interval between video publishing and collection It also has a significant impact on the number of reviews on travel intention; S6.3. Number of followers of the publisher 、Number of video likes , Number of video comments 、Number of video shares And the number of video collections It also has a significant impact on the number of comments on travel intention.

[0027] Preferably, before applying the multiple linear regression model, the key variables , , , Logarithmic transformation and standardization were performed, and each variable was added with 1 and then the natural logarithm was taken. The variables were then centered and scaled to unit variance to reduce the skewness and scaling differences of the key variable data and enhance statistical efficiency and explanatory power.

[0028] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0029] 1. The user travel intention quantification method of the present invention improves the scientificity and effectiveness of tourism marketing strategies by combining machine learning image classification technology with econometric regression analysis; reduces manual intervention through automated image classification methods, improves the objectivity and accuracy of data processing, and ensures the consistency and repeatability of results; and objectively evaluates how each element affects the audience's travel intention through precise analysis of key visual elements in short videos - portraits and scenery, helping tourism marketers to more effectively design attractive marketing content for target audiences.

[0030] 2. Through the application of the user travel intention quantification method of the present invention, the attractiveness of tourist destinations can be enhanced, and by providing scientific decision-making support, the sustainable development and innovation of the tourism industry can be promoted. A new and more objective evaluation and optimization method can be provided for tourism marketing, which is of great value to improving the market performance of tourist destinations and brands. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flow chart of a method for quantifying user travel intention based on the presentation style of short video visual content of the present invention; Figure 2 It is a schematic diagram of key frames of a method for quantifying user travel intention based on the presentation style of short video visual content of the present invention; Figure 3 is a three-level encoding result diagram provided by an embodiment of the present invention; Figure 4 It is a structural diagram of the Inception-v3 model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the application is further elaborated in detail below in conjunction with the accompanying drawings. The described embodiments are only a part of the embodiments involved in the present invention. All non-innovative embodiments of other researchers in the field on this embodiment belong to the protection scope of the present invention. At the same time, for the step numbering in the embodiment of the present invention, it is only set for the convenience of explanation, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0033] In one embodiment of the present invention, a method for quantifying user travel intention based on the presentation style of short video visual content is used to complete the image classification of short videos and further quantify the user's travel intention, such as Figure 1 As shown, the following steps are included:

[0034] Step S1, data collection and image preprocessing: collect tourism short video samples, including comments, likes, favorites, shares, fans, video length and release time, etc., to construct a data set.

[0035] Specifically, in this embodiment, a purposeful sampling technique was used to select keywords such as "beach", "island", "Yalong Bay", "Nan'ao Island" and "Dongshan Island" to search for short videos related to seaside scenery categories and coastal city attractions. A total of 1,941 short video data published between January 1 and July 17, 2024 were crawled using the Python program. The collected data includes user name, number of fans, publishing time, video length, video description, number of likes, number of shares, number of comments, number of favorites, and video links.

[0036] Since the research focus is on video content analysis, all text-based videos containing only static images are excluded in this embodiment. In addition, the collected data are strictly preprocessed, including deleting duplicates and correcting missing values.

[0037] Furthermore, in addition to the video data, this embodiment also batch crawled the comment data of each video (excluding sub-comments), and collected a total of 676,500 comments to ensure data integrity and high accuracy of analysis, providing a solid data foundation for the subsequent quantification of user travel intentions.

[0038] Step S2, key frame extraction and image preprocessing: Apply the inter-frame difference method to the downloaded short videos to extract local maxima, which represent the key frames of each video. Then, 15% of the images are randomly selected from these key frames for manual annotation for subsequent analysis.

[0039] First, the downloaded short video is segmented into frame sequences frame by frame; then, the pixel difference between each frame and its previous frame is calculated to identify significant changes between frames.

[0040] In order to reduce noise interference and improve the accuracy of detection, the moving average method is used to smooth the difference between frames. Figure 2 As shown, the local maximum method is used to detect key change points in the smoothed difference data, so as to accurately extract the key frames in the video.

[0041] Specifically, in this embodiment, after processing 1941 videos, a total of 23693 key frames are extracted, with an average of about 12 key frames extracted from each video. The specific number is adjusted according to the complexity of the video content and the frequency of scene changes.

[0042] Subsequently, 15% of these key frames are randomly selected for manual annotation to prepare a standardized data set for subsequent video content analysis and machine learning training. The manually annotated key frames will be loaded and subjected to necessary preprocessing operations, including but not limited to: color correction, resizing, and contrast enhancement, to ensure that the data quality meets the requirements of subsequent processing.

[0043] Step S3, travel intention coding: For the video comments collected in step S1, 5% are randomly selected for three-level coding, and the travel intention comments are divided into three categories: expressing interest and plans, asking for travel destination information, and interactive behavior. The number of travel intention comments in all video samples is counted through keyword matching method.

[0044] In the data coding and analysis process of this embodiment, a systematic coding method is adopted to ensure the consistency and reliability of data processing.

[0045] First, 20,000 reviews were randomly selected from the 676,500 crawled reviews for manual annotation, and the reviews reflecting travel intentions were screened and preliminarily coded.

[0046] Then, if Figure 3 As shown in the figure, the content of tourism intention comments is divided into types through a three-level coding method: In the open coding stage, NVivo 11.0 software was used to read the comments word by word and repeatedly to conceptualize and categorize the original data to ensure a coding process without subjective bias. To improve the reliability of coding, a double coding procedure was adopted, with two coders annotating the data separately and comparing the coding results. It was found that 95% of the coding results were consistent. For the inconsistent parts, after discussion and consensus, the initial concepts were finally classified into 83 categories.

[0047] Subsequently, the axial coding phase was entered to logically classify and merge categories with similar and related meanings, forming 20 supercategories.

[0048] In the selective coding stage, the supercategories formed during the axial coding process were further focused and aggregated into 12 core categories: strong intention, exploratory interest, specific plan, action preparation, play activities, location inquiry, contact information, cost information, travel preparation, transportation route, finding companions, and participation in social activities.

[0049] On this basis, we conducted in-depth research and logical integration on the comments in the short video comment area, and classified the comments involving users expressing their interest in a certain tourist destination or their future travel plans into the "expression of interest and plans" category; the behavior of seeking information such as the location, cost, and strategies of a specific tourist destination into the "asking for information about the tourist destination" category; and the behavior of interacting with others, such as finding companions and participating in social activities, was defined as the "interactive behavior" category.

[0050] Through this process, short video comments containing travel intentions were classified into three main categories: "expression of interest and plan", "asking for information about travel destinations", and "interactive behavior". These three categories not only effectively summarize the core aspects of users' travel intentions, but also reflect the internal connections and importance between the core categories.

[0051] Finally, the number of travel intention comments of all samples in the sample data set is counted by keyword matching method. The specific method is as follows: First, keyword selection based on coding categories was performed.

[0052] In this embodiment, after completing the selective coding, three main categories are determined: "expressing interest and plan", "asking for information about tourist destinations", and "interactive behavior". For each category, keywords that can represent its core content are selected.

[0053] For example, keywords such as "want to go", "go go go", and "plan to go" are used to mark the "express interest and plan" category; keywords such as "where", "location", and "how to get there" are used to mark the "ask for information about tourist destinations" category; keywords such as "touch up" and "check in" are used to mark the "interactive behavior" category.

[0054] Then, perform keyword matching.

[0055] In this embodiment, keyword matching is performed on all comments in the sample data set, and each comment containing any of the above keywords or a combination of multiple keywords is marked as a comment containing a specific travel intention. This matching method ensures that all relevant comments can be fully and accurately identified.

[0056] Finally, statistical analysis was performed.

[0057] In this embodiment, a statistical analysis is performed on the number of comments containing travel intentions in a single short video. By performing a quantitative analysis on the travel intention comments in each short video, it is possible to not only understand the distribution of users' interests in different tourist destinations, but also to further explore the specific concerns and behavior patterns of users when expressing their travel intentions.

[0058] Step S4, Inception-v3 model training and evaluation: Divide the preprocessed image data into a training set (64%), a validation set (16%), and a test set (20%); adjust the pre-trained Inception-v3 model, remove the top layer, add a global average pooling layer, a custom fully connected layer, and an output layer; use the validation set to evaluate the performance during training, and adjust the model parameters according to the performance chart to optimize the effect; finally, verify the classification accuracy and generalization ability of the model through the test set.

[0059] Specifically, in this embodiment, step S4 includes the following sub-steps:

[0060] Step S4.1: For the 3169 manually labeled image samples, 2028 of them are divided into training sets, 507 into validation sets, and 634 into test sets.

[0061] Step S4.2: Figure 4 As shown in Figure 1, the pre-trained InceptionV3 model is adjusted to suit the image classification task.

[0062] First, load the InceptionV3 model without the top layer and set its base layer to be non-trainable to retain its efficient feature extraction capabilities. With this strategy, you can focus on customizing the top of the model without retraining the entire network.

[0063] Specifically, a global average pooling layer (GlobalAveragePooling2D) is added to the original InceptionV3 model architecture to compress the feature map into a vector of fixed length.

[0064] Subsequently, a fully connected layer is introduced to further process these feature vectors, and finally a binary classification task is achieved through an output layer using a sigmoid activation function.

[0065] Next, during the model compilation phase, the Adam optimizer and binary cross entropy loss function were used to optimize the classification performance. To avoid overfitting, this implementation also used the Early Stopping method.

[0066] In addition, in order to enhance the generalization ability of the model and solve the problem of data imbalance, this embodiment also introduces data enhancement technology and category weight adjustment strategy. Through these measures, the improved InceptionV3 model can not only effectively extract high-level features from the input image, but also show good performance and adaptability in practical applications.

[0067] In this example, the improved InceptionV3 model starts with an input layer of size 299×299×3, which is suitable for RGB images. The subsequent data flow enters a series of complex base layers (Blocks) consisting of convolutional layers, pooling layers, and concatenation layers. The base layers are designed to gradually extract and combine feature maps of different scales.

[0068] For example, Block 1 contains multiple modules, each of which contains 1×1, 3×3, 5×5 convolution kernels and a maximum pooling layer, and the results of these operations are merged in the concatenation layer. This design allows the model to capture both fine-grained and coarse-grained feature information.

[0069] The specific module details are as follows: Block 1: This block contains three modules, each of which first uses 1×1 convolution to reduce the number of channels, and then applies 3×3 and 5×5 convolutions to expand the receptive field respectively, while combining the maximum pooling layer to capture spatial information at different levels.

[0070] Block 2: A further deepened network consisting of five modules adopts a similar but more complex pattern and adds 7×7 convolution kernels to capture a wider range of contextual information.

[0071] Block 3: The last few modules continue this process, but are simplified in architecture and focus on refined feature extraction.

[0072] The operations at each layer change the data dimension. For example, after multiple convolutions and pooling, the original input data gradually shrinks and deepens its feature representation capabilities.

[0073] After the data flows through the base layer, it is passed to the global average pooling layer. This step compresses the entire spatial dimension of each feature map into a single value, that is, for each feature map, the average of all its pixel values ​​is calculated.

[0074] In this way, no matter what the size of the input image is, the output will be a vector of fixed size, which in this embodiment is 8 × 8 × 2048. This process not only reduces the number of parameters, but also effectively prevents overfitting.

[0075] Next, this 8×8×2048 vector is fed into one or more fully connected layers for further processing in order to learn higher-level feature representations. Finally, a binary classification task is performed through an output layer with a sigmoid activation function to produce the final prediction result.

[0076] Through the above process, the improved InceptionV3 model can efficiently process the input image, gradually extract useful feature information, and finally make accurate classification decisions based on these features. This design not only improves the performance on specific tasks, but also enhances the adaptability of the model to various application scenarios.

[0077] Step S4.3: In order to select an image classification model suitable for step S4, mainstream image classification models such as ResNet50, InceptionV3, VGG16, VGG19 and DenseNet121 were comprehensively tested and their performance was compared.

[0078] The classification performance of each model was comprehensively evaluated by generating performance charts such as confusion matrix, loss rate curve, precision curve, accuracy curve and ROC curve. To prevent overfitting, early stopping, regularization and data enhancement techniques were uniformly used in the training process to improve the generalization ability of the model.

[0079] After comprehensive performance evaluation, this embodiment finally determines the fine-tuned InceptionV3 as the main image classification model. In order to adapt to the processing requirements of different models, the main adjustments include the size of the input picture. The learning rate of all models is uniformly set to 0.0001 to ensure the stability and efficiency of the training process. In addition, all models use a batch size of 32, which is a balanced choice after considering memory efficiency and model performance. The optimizer used is Adam, which makes the training more stable due to its adaptive learning rate capability. The loss function is binary_crossentropy, which is suitable for binary classification problems and can effectively process category outputs.

[0080] Furthermore, this embodiment monitors the loss of the validation set through the Early Stopping method, and the patience parameter is set to 10, aiming to prevent overfitting and reduce unnecessary training time. Data enhancement includes random rotation, scaling, flipping and other operations, which enhances the generalization ability of the model when facing new and unseen samples. Through these detailed configurations and parameter optimizations, the scientificity and applicability of the model selection are guaranteed, so that the model can achieve better performance in practical applications.

[0081] Table 1 further shows the detailed measurement results of the performance of different machine learning models. For a given test set, the calculation formulas for precision, recall and F1 score are: ; ; ; Among them, TP, FP and FN represent true positive, false positive and false negative, respectively.

[0082] Table 1: Machine Learning Model Performance

[0083] Step S5, image classification and style judgment: Input the image to be classified into the trained Inception-v3 model, and the model automatically extracts image features and classifies them. Each classification outputs a binary result, "0" or "1", indicating that the image does not contain or contains a specific target element (such as scenery or portraits). Based on the model classification results, the visual presentation style of the image is judged.

[0084] In this example, the image classification and style judgment process is implemented by using the trained InceptionV3 model. First, the image to be classified is input into the model, where the model uses its deep learning architecture to automatically extract the key features in the image. Then, the model classifies according to these features, and the output of each classification is a binary result, "0" or "1", which represents whether the image contains a specific target element. For example, "0" may mean that there is no portrait but only scenery, while "1" means that the image contains a portrait.

[0085] This classification mechanism allows the system to determine the visual presentation style of an image in an efficient and automated manner. For example, based on the presence and proportion of portrait and landscape elements in the image, the system can determine whether the image is more inclined to a person-centric style or a landscape-centric style. The method of this embodiment is extremely valuable in a variety of application scenarios, such as automatic content classification, augmented reality applications, or personalized media recommendations.

[0086] Step S6, quantitative analysis of impact: Based on the image classification results and the number of travel intention comments in the video, a multivariate linear regression model is constructed to quantitatively analyze the impact of the visual presentation style and other video features of the short video on the user's travel intention.

[0087] Specifically, this embodiment studies the influence of the visual presentation style of short videos on the user's travel intention by establishing a generalized linear model Poisson regression, which is as follows:

[0088] ; ; ; ; In the formula, The number of comments about users’ travel intentions, is the number of landscape-based pictures in the video keyframes, is the number of pictures with portraits as the main feature in the key frames of the video, is the video duration, The time interval from video publishing to video collection. is the number of followers of the publisher, Number of likes for the video, is the number of video comments, is the number of video shares, is the number of video collections; β0~β8 are regression parameters, and ε is a random interference term.

[0089] Preferably, before applying the generalized linear model, , , , Several key variables such as , are logarithmically transformed and then standardized to reduce the skewness and scaling differences of the data and increase its statistical power and explanatory power in subsequent models. The specific operation is to add 1 to each variable and take the natural logarithm, then center and scale it to unit variance, so that the data is closer to normal distribution, which is particularly important for the application of regression models.

[0090] Then, the model was applied to explore the impact of short video visual content presentation style on users’ travel intention. The results are shown in Table 2.

[0091] Table 2. The impact of short video visual content presentation style on user travel intention

[0092] *p<0.05; **p<0.01, ***p<0.001

[0093] Specifically, first, the regression results of Model 1 and Model 2 are analyzed.

[0094] Regression analysis shows that the number of key frames dominated by portraits in the video It has a significant negative impact on users’ travel intention. Specifically, as the number of portrait images in the video increases, the number of viewers’ travel intention comments decreases significantly (regression parameter , , Model 1).

[0095] The regression parameter β indicates the degree of influence of the independent variable on the dependent variable. For example, in Model 1, β1 indicates that for every additional key frame dominated by portraits, the number of comments on the audience's travel intention decreases by 0.024 on average.

[0096] The p value is used to measure statistical significance. In this example, p < 0.01 and p < 0.05 mean that the observed results are statistically significant at the 1% and 5% significance levels, respectively. For p < 0.01, the null hypothesis can be rejected at a 99% confidence level, while for p < 0.05, the null hypothesis is rejected at a 95% confidence level, which means that there is enough evidence to reject the null hypothesis and believe that the number of key frames has a significant effect on travel intention.

[0097] This result suggests that when videos contain more people-related elements, it may distract viewers from tourist destinations or attractions, thereby reducing their interest in traveling.

[0098] In contrast to the number of key frames dominated by portraits, the number of key frames dominated by scenery in the video The regression results show that as more natural landscapes or beautiful attractions are shown in the video, the audience's willingness to travel increases (regression parameter , , Model 2). This may be because natural landscapes themselves have strong appeal and can stimulate the audience's desire to explore. Especially when it comes to tourist destinations, rich scenery pictures can effectively enhance the audience's interest.

[0099] When considering the square term of the number of key frames dominated by portraits When , the regression coefficient is negative and significant (the regression parameter , , Model 1), which shows that as the number of portrait images increases, their negative impact on travel intention increases. Too many portrait images may cause visual fatigue to the audience, thereby reducing their interest in travel content. Therefore, although a moderate amount of human elements may have a certain effect on emotional resonance, when their number is too large, it will weaken the audience's willingness to travel.

[0100] Similar to the number of key frames dominated by portraits and the square of the number of key frames dominated by scenery ) also showed a significant negative impact (regression parameter , , Model 2). This means that although increasing the number of scenery pictures can increase the audience's interest in traveling, when there are too many scenery pictures, the audience's interest may begin to decline. Excessive display of natural landscapes may cause visual fatigue or information overload, thereby weakening its positive impact on travel intention, reminding video producers to grasp the degree of scenery display.

[0101] For video duration The regression coefficients are different in different models. When the number of key frames dominated by portraits and their square terms are used as independent variables (Model 1), the video duration The regression coefficient is , and in This result shows that the increase in video length will promote the increase in the number of comments on travel intention. Longer videos can provide more information and details, especially in videos with people as the main feature, the audience tends to be more emotionally resonant. If the characters in the video show emotions such as pleasure and happiness, it may infect the audience and stimulate their interest in travel. Longer duration can further strengthen the transmission of this emotion, and the audience is more likely to have more travel intention and leave comments during the viewing process. Therefore, in videos with people as the main feature, longer duration helps to increase the audience's participation and willingness to comment.

[0102] When the number of key frames dominated by scenes and their square terms are used as independent variables (Model 2), the regression coefficient of video duration is , and in This result indicates that increasing the length of the video will reduce the number of travel intention comments. Scenery-oriented videos usually focus on showing the beauty and practical value of the scenic spot (for example, whether the scenic spot is worth visiting), but if the quality of the scenery is not high or the scenic spot itself lacks sufficient appeal, longer videos may expose these shortcomings. For scenery videos, viewers usually expect to see more visually impactful and creative content, and longer duration may make the shortcomings of the scenic spot more obvious, resulting in a decrease in audience interest, thereby reducing comments and interactions. Therefore, if the scenery-oriented video is too long, it may cause audience fatigue and reduce their participation.

[0103] The time interval between video publishing and video collection ) has significant positive and negative differences in the regression coefficients in different models. When the number of key frames dominated by portraits and its square term are used as independent variables (Model 1), the regression coefficient is , and in , indicating that a longer time interval has a negative impact on the number of comments on travel intention. This may be because a longer time interval means that the audience's attention and participation may decrease after the video is released, resulting in a decrease in interactivity. Especially when it comes to human scenes, a long time interval may gradually reduce the audience's interest, thereby reducing the number of comments. On the other hand, when the number of key frames dominated by scenery and its square term are used as independent variables (Model 2), the regression coefficient is , and in This may indicate that a longer time interval can increase the sedimentation effect of video content, allowing the video to regain the audience's attention after a period of time, especially when it comes to content related to scenery, the continuous exposure of the video may attract more viewers to participate in the comment.

[0104] In addition to these independent variables and the above two control variables, the regression coefficients of the remaining control variables in the four models show consistent significant trends. , video sharing quantity and video collection quantity The regression coefficients are all positive and significant.

[0105] The positive relationship between the number of video shares and the number of travel intention comments indicates that when viewers actively share videos, they not only recognize the video content to a certain extent, but may also influence others in their social circles and networks. When a video is widely shared, it will bring higher exposure and attract more potential viewers to watch and comment on the video. The act of sharing itself also shows that the audience has a high interest and engagement in the video content, which further motivates them to leave comments, especially when it comes to travel destinations. Sharing behavior can deepen the audience's interest in the destination and prompt them to express their personal opinions.

[0106] An increase in the number of video comments is usually a reflection of active interaction, and may also drive more viewers to comment. People often develop a certain sense of social identity or curiosity after seeing other viewers leave comments, which will also stimulate their further interest in the video content, especially in travel-related videos, where viewers share their opinions, travel experiences, or ask related questions in their comments. This kind of interaction enhances the sociality and participation of the video content, which in turn increases the willingness of other viewers to comment, forming a virtuous feedback loop.

[0107] An increase in the number of video collections indicates that viewers have a high level of interest in and identification with the video content. Collection behavior usually means that viewers want to review or refer to the video in the future, indicating that they think the content of the video is valuable or relevant to their personal needs. In travel videos, viewers may collect content involving destinations or travel advice that they are interested in so that they can view it again or plan a trip in the future. An increase in the number of video collections not only reflects the audience's long-term interest in the content, but also encourages them to share, discuss, or actively participate in more interactions with other viewers in the future, thereby increasing the number of comments indicating travel intention.

[0108] The number of followers of the publisher and video likes The regression coefficients of are all negative and significant, indicating that the increase of these factors is negatively correlated with the number of comments on travel intention. One possible reason is that when a scenic spot is too popular, the audience may hesitate to go there, worrying that the crowds will affect the travel experience, which may lead to a decrease in their interest in tourism-related content, thereby reducing the number of comments on travel intention. In addition, the large number of fans of the blogger may mean that the video has involved a large number of audience groups during the dissemination process, which may also make the atmosphere of short video tourism marketing more intense, further pushing up the audience's expectations. For some viewers, overly "commercial" marketing may cause disgust or "dissuade" effects, resulting in a decrease in their participation and willingness to interact. Therefore, although the number of fans and likes can increase the exposure of the video, the increase of these factors may have a negative relationship with the audience's travel intention under certain conditions.

[0109] For Model 3 and Model 4, the main focus is on studying the independent impact of the two variables and their nonlinear relationship.

[0110] In model 3, the regression results show that the number of key frames dominated by scenery in the video And the number of key frames mainly for portraits Both have a significant positive impact on the number of comments on visitors' travel intentions.

[0111] Specifically, The regression coefficient is 0.290, which means that for every additional key frame dominated by scenery, the number of comments on travel intention will increase by about 0.290. This shows that scenery elements have a strong appeal to the audience's travel intention and can effectively stimulate the audience's interest and desire to comment. The regression coefficient The value is 0.105, that is, for every additional key frame with a portrait as the main feature, the number of comments on travel intention increases by 0.105. This shows that although portrait images also have a certain positive impact on travel intention, their influence is smaller than that of scenery.

[0112] In addition, the time interval between video publishing and video collection The regression coefficient The value is 0.031, which is significant at the 1% significance level, indicating that a longer time interval is associated with a greater number of travel intention comments. This may be because the video continues to be exposed after it is released, which can attract the audience's attention and motivate them to leave comments.

[0113] The video duration The regression coefficient The value is -0.215, indicating that there is a negative relationship between longer video length and the number of travel intention comments. This may mean that overly long videos will cause viewers to lose attention and affect their willingness to interact.

[0114] In Model 4, in addition to adding and In addition, we also considered and , that is, the square term of the number of key frames is introduced. The regression results show that the introduction of the square term further reveals the nonlinear effect of the scenery and portrait elements in the video on the audience's willingness to travel.

[0115] Specifically, The regression coefficient is 0.277, and The regression coefficient It is -0.014, which indicates that the impact of the number of key frames dominated by scenery on tourism intention shows a decreasing trend, that is, when the number of scenery pictures exceeds a certain number, its positive effect will weaken and may even bring negative effects. The regression coefficient is 0.117, and The regression coefficient It is -0.015, which also shows that the impact of the number of key frames dominated by portraits on travel intention is also nonlinear. As the number of portrait frames increases, the positive effect gradually weakens and may even bring negative effects.

[0116] In summary, according to the regression results of Model 1 and Model 2, the composition of video content, especially the display of portraits and scenery, significantly affects the audience's willingness to travel. The impact of video length is closely related to the nature of the content: in videos dominated by people, longer duration helps to enhance emotional resonance and improve travel intention; while in scenery videos, too long duration may expose the shortcomings of the scenic spots and reduce the audience's interest. The effect of the time interval between video release and collection on the number of comments varies depending on the content type: the number of comments for videos dominated by portraits decreases after a longer time interval, while the number of comments for scenery videos increases after a longer time interval, indicating that the accumulation effect of video type and release time affects interactivity. The sharing, commenting and collection of videos reflect the audience's high level of participation in the content and promote the increase in the number of comments on travel intention. Although the number of followers and likes of the publisher are helpful for video exposure, they do not determine the success of tourism marketing. Therefore, creators should focus on improving the quality of video content, even if starting from scratch, which will more effectively attract accurate audiences, increase interaction and comments, and thus improve travel intention.

[0117] In contrast, the regression results of Model 3 and Model 4 reveal the complexity of video content, especially the diminishing effect of portrait and scenery elements. In Model 3, when the number of key frames dominated by portraits and the number of key frames dominated by scenery are considered together, the positive impact of portrait elements on travel intention is more significant, while the impact of scenery elements is still positive, indicating the relative importance of the two. Model 4 further introduces the square terms of portraits and scenery, and finds that the increase in the number of portrait and scenery elements does not have a linear effect on travel intention, but has a diminishing effect. This shows that in the process of short video creation, moderately increasing portrait and scenery elements can help increase the audience's interest, but exceeding a certain amount may cause visual fatigue, which in turn reduces travel intention. Therefore, properly balancing the number of video content elements and controlling the video length and release time interval can significantly improve the audience's interactivity and travel intention.

[0118] In summary, this paper proposes a method to quantify users' travel intention based on the presentation style of short video visual content, and verifies its effectiveness in short video tourism marketing. In particular, by optimizing the ratio of portraits to scenery in the video, this method can effectively promote users' travel intention, which has important theoretical value and practical application significance.

[0119] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for quantifying user travel intention based on the presentation style of short video visual content, characterized in that: The steps include: Step S1, data collection and image preprocessing: collect tourism short video samples and perform image preprocessing to build a sample data set; Step S2, key frame extraction and labeling: for the short video data in the sample data set, the local maximum is extracted by the frame difference method as the key frame of each short video, and some data is randomly selected from the key frame for labeling; Step S3, tourism intention coding: based on the comment information of short videos in the sample data set, randomly select some data, classify tourism intention through the three-level coding method, and count the number of tourism intention comments of all samples in the sample data set through the keyword matching method; Step S4, Inception-v3 model training and evaluation: Divide the preprocessed image data into a training set, a validation set, and a test set, adjust the pretrained Inception-v3 model, remove the top layer, add a global average pooling layer, a custom fully connected layer, and an output layer, use the validation set to evaluate the performance during training, adjust the Inception-v3 model parameters according to the performance chart, and verify the classification accuracy and generalization ability of the Inception-v3 model through the test set; Step S5, image classification and style judgment: input the image to be classified into the Inception-v3 model trained in step S4, automatically extract image features and classify them, and output a binary result for each classification, indicating that the image does not contain or contains the target element. Based on the classification result of the Inception-v3 model, judge the visual presentation style of the image; Step S6, quantitative analysis of impact: Based on the image classification results obtained in step S5 and the number of travel intention comments in the video obtained in step S3, a multivariate linear regression model is constructed to quantitatively analyze the visual presentation style of the short video and its impact on the user's travel intention.

2. The method for quantifying user travel intention based on short video visual content presentation style according to claim 1 is characterized in that: In step S1, the travel short video sample includes: comments, number of likes, number of favorites, number of shares, number of fans, video length and release time information; the image preprocessing includes: deleting duplicate data, correcting missing values, and removing text and image videos.

3. The method for quantifying user travel intention based on short video visual content presentation style according to claim 1 is characterized in that: In step S2, the key frame extraction and labeling includes the following sub-steps: S2.1, split the short video data in the sample data set into frame sequences frame by frame, calculate the pixel difference between each frame and the previous frame, and identify the significant changes between frames; S2.2, using the moving average method to smooth the difference between frames, using the local maximum method to detect key change points in the smoothed difference data, and extracting the key frames in the video; S2.

3. Randomly select part of the data from the key frames according to a preset ratio, perform manual annotation, form a standardized data set, load the key frame data in the standardized data set and perform image data processing, including: color correction, size adjustment, contrast enhancement, and set the image resolution to a uniform size, maintain the color format, and scale the image pixel value to between 0 and 1.

4. The method for quantifying user travel intention based on short video visual content presentation style according to claim 3 is characterized in that: The three-level coding method in step S3 includes: open coding, main axis coding, and selective coding; the tourism intention classification includes: expressing interest and plan, asking for information about tourist destinations, and interactive behavior.

5. The method for quantifying user travel intention based on short video visual content presentation style according to claim 4 is characterized in that: The tourism intention coding in step S3 includes the following sub-steps: S3.

1. Based on the comment information of short videos in the labeled sample data set, randomly select some data, screen out the comments that reflect travel intentions, and perform preliminary coding; S3.

2. Classify the content of travel intention comments by three-level coding method: S3.2.

1. In the open coding stage, the original review information was read word by word using software to conceptualize the initial concepts. The double coding method was used to mark and code them respectively. After comparing and unifying the coding results, the initial concepts were classified. S3.2.

2. In the axial coding stage, categories with similar and related meanings were logically classified and merged to form supercategories; S3.2.

3. In the selective coding stage, the supercategories formed in the main axis coding stage were aggregated into core categories, including: strong intention, exploration interest, specific plan, action preparation, play activities, location inquiry, contact information, cost information, travel preparation, transportation route, finding companions and social activity participation; S3.2.

4. Summarize the core of users’ travel intentions: Integrate the comments in the short video comment area and classify the short video comments containing travel intentions; S3.

3. Count the number of travel intention comments of all samples in the sample data set by keyword matching method: S3.3.

1. Keyword selection based on coding categories: Based on the classification results of short video comments containing travel intentions in step S3.2.4, select several keywords representing the core content of each category; S3.3.2, Keyword matching: Keyword matching is performed in all comments in the sample data set, and each comment containing one or more keywords is marked as a comment containing travel intention; S3.3.

3. Statistical analysis: Conduct statistical analysis on the number of comments containing travel intentions in a single short video, quantify the distribution of users’ interests in different tourist destinations, and the specific concerns and behavior patterns of users when expressing their travel intentions.

6. The method for quantifying user travel intention based on short video visual content presentation style according to claim 5 is characterized in that: In step S3.2.4, the short video comments containing travel intention are classified into: expressing interest and plan, asking for travel destination information, and interactive behavior; Comments involving users expressing interest in a travel destination or future travel intentions were classified as expressed interest and plans; Comments seeking information about the location, cost, and guide of a tourist destination were classified as asking for information about the tourist destination; Comments about seeking companions, participating in social activities, and interacting with others were classified as interactive behaviors.

7. The method for quantifying user travel intention based on short video visual content presentation style according to claim 5 is characterized in that: In step S4, the pre-trained Inception-v3 model is adjusted, including the following sub-steps: S4.

1. Based on the pre-trained Inception-v3 model, load the model without the top layer and set the base layer to be non-trainable. S4.

2. Add a custom output layer, including: a global average pooling layer and a fully connected layer, wherein the global average pooling layer is used to compress the feature map into a feature vector of fixed length, and the fully connected layer is used to process the feature vector and perform binary classification through an output layer using a sigmoid activation function; S4.

3. Compile using the Adam optimizer and binary cross entropy loss function, using early stopping to prevent overfitting. S4.

4. Enhance the generalization ability of the Inception-v3 model through data augmentation and category weight adjustment.

8. The method for quantifying user travel intention based on short video visual content presentation style according to claim 7 is characterized in that: In step S5, the Inception-v3 model performs classification based on image features, and outputs a binary result of 0 or 1 for each classification, indicating that the image does not contain or contains the target element, respectively; The visual presentation style includes: experience style and landscape style. If the target element of the image is mainly a close-up of a portrait, it is classified as the experience style; If the target element of the image is mainly landscape display, it is classified as landscape style.

9. The method for quantifying user travel intention based on short video visual content presentation style according to claim 8 is characterized in that: In step S6, the multivariate linear regression model is a generalized linear model Poisson regression, and the variables of the multivariate linear regression model include: the number of comments on user travel intention , the number of pictures with mainly landscape in the video keyframes , the number of images with portraits as the main feature in the video keyframes , video duration The time interval from video publishing to video collection , the number of followers of the publisher , video likes , the number of video comments , video sharing quantity , video collection number ; The generalized linear model Poisson regression is expressed as follows: ； ； ； ； In the formula, Indicates a and The vector of β 0 ~ β 8 is the regression parameter, ε is the random interference term; The relationship between variables is analyzed by multivariate regression model, and the regression parameters are β 0 ~ β 8 Quantify the degree of influence of variables on the number of comments on tourism intention respectively; by determining the setting principles of regression parameters and statistical significance, test and determine which variables have a significant impact on the number of comments on tourism intention.

10. The method for quantifying user travel intention based on short video visual content presentation style according to claim 9, characterized in that: Before applying the multiple linear regression model, the key variables , , , Logarithmic transformation and standardization were performed, and each variable was added with 1 and then the natural logarithm was taken. The variables were then centered and scaled to unit variance to reduce the skewness and scaling differences of the key variable data.

Images