Integration of Chu Culture Elements and Big Data Analysis Strategies in the Digital Development of Sports Culture

The digitization of sports culture can be divided into “sports culture” and “digitization” according to the key words. “Culture” is the most important and ambiguous concept in human society, and according to the important deployment of the national cultural digitization strategy, sports culture is an important field of digital construction, and the definition of sports culture digitization should follow the general requirements of the concept of digitization [17-18]. Thus, integrating the definitions of sports culture and digitalization, the digitalization of sports culture is defined as the progressive process of changing the production mode, product quality, service process and governance capacity of sports culture with the help of the computing, communication, linking and application functions of digital technologies such as 5G, Internet of Things, big data, cloud computing, and so on, so as to enrich the form of its presentation, revitalize the vitality of the culture in its time, and enhance the efficiency of cultural dissemination.

Digitalization of sports culture production methods, traditional sports culture is mainly processed and produced by means of on-site display, film and television shooting, etc., which makes it difficult to display the connotation and qualities of sports culture. The digital production method can facilitate the integration of sports culture and digital technology, leading to intelligent creation, customized production, and digital management. For example, the NFT digital collection derived from blockchain technology further promotes the generation of new sports culture businesses in the sports clothing industry, sports events, and so on. Sports intangible cultural heritage can be reproduced in bulk through 3D printing technology, which is based on digital model files and offers services such as personalized customization and creative DIY design. Thirdly, the supply of sports and cultural products is digitalized. Traditional sports and cultural products display cultural products in the form of physical objects, pictures, videos and other forms, and with the passage of time can lead to the lack of sports and cultural information, which in turn affects the quality of dissemination.

Chu culture is a cultural phenomenon developed in a specific region by a specific ethnic group. In terms of ethnicity, the Chu nation is the collective name for the generation of races in the Jianghan Plain of the Yangtze River Basin. From the cultural name to talk about, is the existing archaeological excavations found since the Zhou Dynasty, to the Spring and Autumn and Warring States, since the Han Dynasty shows the cultural characteristics of the independent region. Therefore, Chu culture in a broad sense is to include all of the above four concepts. In a narrow sense, Chu culture refers to the cultural forms created by the working people in the middle reaches of the Yangtze River during the Chu period. This culture includes not only sorcery and myths, poetry and prose, but also folklore, design, and artifacts.

Chu artifacts represented by lacquerware are a unity of typical forms, patterns, and colors of the elements of Chu culture. “Unique shape, exquisite decoration, mysterious color” is a summary of its characteristics. Mainly in: the shape of phoenix birds, beasts, human-animal form combinations, etc. As the theme, the surface of the natural lacquer is coated with red and black as the main color, with patterns such as cloud patterns, phoenix patterns, and taotie patterns. The combination of beasts and human-animal forms represents ferocity and strength, reflecting the self-reliance of the Chu people in their struggle with nature and the vassal states. The red and black color represent passion and tolerance, and it is the embodiment of the romanticism of the Phoenix Nirvana, the rebirth of fire, and the combination of human and animal forms. In short, whether it is the unique shape and mysterious color of Chu lacquer ware, or Chu silk weaving pattern of phoenix birds and auspicious clouds, are the Chu people in the long-term reality of production, life, self-improvement, openness and tolerance and other spiritual connotations of the vivid embodiment. The typical forms, patterns, and colors of the Chu culture together construct a vibrant, romantic, and charming world of imagery.

The residual learning unit diagram is shown in Figure 1, ResNet uses residuals to solve the problem of degradation of information, adding a participatory layer on the basis of the neural network structure to learn the residuals between the input and output of information, and after learning, it can improve the convergence speed of the model when it is running, and also improve the accuracy rate, in the structure of the residual neural network, the outputs of the n-layers are connected together by using the summing, which can easily result in the The propagation of information has an impact on the gradient disappearance problem, in order to make up for the shallow network can only extract the edges of the image and color, etc., to improve the accuracy of recognition, and then deepen the number of network layers to extract the main features of the image, the residual network is composed of many residual blocks, a residual network is a residual network is a lot of residual blocks stacked on top of one another to form a deep learning neural network, which is used for the input of each layer as a benchmark and continuously learns to form a residual function [19-20]. This residual function is easier to tune and can continuously increase the number of layers in the network to form a multilayer network structure.

Figure 1.

The study unit process of the residual study

The output of layer n–1 of a conventional convolutional neural network is the input to layer n: (1) Xn=Hn(Xn−1)

With the inclusion of a residual structure, the output of the next layer is affected by the input of the previous layer: (2) Xn=Hn(Xn−1)+Xn−1

The mathematical expression for the residual structure is: (3) xl+1=xl+F(xi,Wl)

After recursion, deep cell L is characterized by: (4) xL=xi+∑i=1L−1F(xi,Wi)

It is shown that any cell L and 1 have the property of residuals between them, so that a cell L of any depth, which is characterized by: (5) xL=x0+∑i=0L−1F(xi,Wi)

Assuming that the loss function is E, the backpropagation can be obtained according to the chain rule as: (6) ∂ε∂xi=∂ε∂xL∂xL∂xi=∂ε∂xL(1+∂∂xi∑i=1L−1F(xi,wi))

Eq. takes everything in the weighting layer into account, and since ∂∂x1∑i=1L−1F(xi,wi) cannot be -1, there is no problem of vanishing gradients.

Fig. 2 Schematic diagram of residual block structure, residual neural network also derives a lot of network structures, such as Res Net18 and Res Net34 and other networks with fewer layers, while the widely used residual networks are Res Net50, Resnet101 and Res Net152 and other networks with more network layers. Take Res Net101 as an example: the first layer is a 7*7*64 convolution, the next layer is 3+4+23+3=33 structural blocks, each structure is 3 layers, so 33*3=99 layers, and finally the fully connected layer is added, so there are 1+1+99=101 layers. The reason for creating a multi-layer network structure is to have other lines that can be connected to the later layers, and then learn the residuals directly.Res Net101 is based on the principle of Res Net, based on the network structure of VGG, and add residual blocks to form the composition of the network, because this network has a total of 101 layers, so it’s called ResNet101, and the sizes of the residual blocks are 1*1, 3*3, and 1*1, respectively. The size of the residual block is 1*1, 3*3, and 1*1 respectively, and the convolutional layers are merged sequentially, and the Relu function is added after each convolutional layer.

Figure 2.

Difference block structure schematic

In this subsection, the algorithm is designed for the feature extraction of cultural elements of Chu culture, migrating multiple convolutional neural networks trained by ImageNet dataset, including InceptionV3, CNN, ResNet, MobileNetV2, DenseNet201, and VGG16. In order to obtain the features with lower dimensions, the two fully-connected layers at the tail of the model are removed and replace all neurons in the output layer of the model to adjust it to 45, taking ResNet as an example, the specific operation steps are as follows.

Overfitting is more likely to occur due to the large number of parameters in the model and the small number of samples of Chu cultural elements images.Stochastic deactivation (Dropout) can cause a neuron’s activation value to stop working with a certain probability when the neural network propagates forward. In order to reduce the occurrence of overfitting phenomenon and improve the efficiency of the model in extracting and recognizing the elements of Chu culture, this subsection conducts experiments on the basis of subsection 2.2.2 of this paper, improves ResNet based on the stochastic deactivation strategy, and names the improved network as ResNet_dropout.The specific operation steps are as follows.

The use of Chu cultural elements should not be limited to a single field, but should involve the integration and sharing of resources through cross-border cooperation. By cooperating with brands or designers from different industries and fields, Chu cultural elements can be introduced to new design concepts and technical means, and applications can be expanded. At the same time, by integrating resources, a more complete industrial chain and market network can be established to promote the development of sports culture at a global level. Through the creative transformation of designers, the elements of Chu culture are integrated into modern design, and new technologies, such as 3D printing and virtual reality, are used to enhance the interactivity and experience of sports cultural and creative products, accurately locate the target consumer groups, and realize the market operation of the products. Combine Chu cultural elements and sports-related creative products to preserve and safeguard traditional culture. Chu cultural elements have their own history and cultural connotation, and the cultural creation should express the “invisible” history in a “tangible” way. The organic integration of Chu cultural elements and sports creative products, under the premise of maintaining and promoting, using modern technology means, give full play to the advantages of Chu cultural elements, maximize the use of different kinds of cultural resources, thus the digital development of sports culture.

We should vigorously integrate the resources of Chu cultural elements, vigorously develop pillar sports and cultural industries, strive to bridge the phenomenon of cultural rupture that has occurred in the course of the development of history and culture, and drive the innovative and digitalized research of local sports culture and its artistic design through the communication path mediated by image symbols. In addition, with the responsibility of enhancing our cultural digestive capacity and innovation ability, we will promote the healthy and rapid development of the cultural economy, actively participate in the competition in the international cultural market, vigorously disseminate the brilliant achievements of Chinese civilization, and realize the diversification, socialization and publicization of the investment main body of the cultural economy. To sum up, only by fully integrating and drawing on the elements of traditional Chu culture, catering to the aesthetic needs of modern people, blending the essence of ancient and modern sports culture, striving to spread sports culture that meets the psychological characteristics of modern people on the basis of integrating the elements of traditional Chu culture, and rejecting the old way of just repeating and revering the past, and inheriting as well as innovating, the quality of digitalization of China’s sports culture can be guaranteed and recognized.

As China’s valuable sports cultural heritage, Chu culture elements are also an important part of the sports culture business that cannot be ignored. Take multicultural integration as a means to promote the development of sports towards digitalization. In line with economic globalization, different cultures in the global society continue to break through the limitations of cultural regions and cultural patterns and go global, gradually transforming many localized elements into resources shared by human culture. With the high degree of development in information technology, people are rapidly exposed to a large number of different cultures from various sources, so the phenomenon of multicultural integration is manifested in almost all fields. In the dissemination of sports culture, this multicultural integration is first reflected in the fusion of sports culture and elements of Chu culture through the opening and closing ceremonies of various large-scale sports events.

The network model training environment is shown in Table 1, and this experiment is conducted using the PyTorch deep learning framework. The experimental configuration is as follows: the GPU is RTX 4060 with 13.8 GB of video memory, the development language is python with version 3.3, the operating system is Ubantu 17.33, the memory is 26.3 GB, the hard disk is 500.00 GB, and the version of Pytorch is 1.10.

Residual Neural Network (ResNet) was trained with the following parameters: 0.0001 learning rate and 200 iterations, respectively. The data preprocessing stage was processed by Gaussian filtering as well as image normalization using the correlation parameters ([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), and the resulting image input was 3 × 224 × 224.

In machine learning, the computational results are usually represented by a confusion matrix, which is shown in Table 2. Precision, recall, accuracy and Fl score are all important metrics for evaluating the model.1 means that the sample is in the positive category, 0 means that the sample is in the negative category, predicted represents the predicted value of the sample, and Actual represents the actual value of the sample. The precision rate, which is the proportion of samples that are truly predicted as positive to all samples that are predicted as positive, is calculated by the formula:

Precision rate, is the proportion of samples that are truly predicted positive to all samples that are predicted positive, calculated by the formula: (7) precision=TPTP+FP

Recall, which is the ratio of samples with a true positive prediction to all true samples, is calculated by the formula: (8) recall=TPTP+FN

Correctness, which is the ratio of correctly categorized samples to all samples, is calculated by the formula: (9) Accuracy=TP+TNTP+FN+TN+FP

The F1 score, which is the formula used to synthesize measures of precision and recall: (10) F1=2TP2TP+FP+FN

Since the samples of each category in the dataset of this paper are 100 images, there is no problem of sample data imbalance, and the accuracy rate is simple to calculate, which can intuitively respond to the goodness of the model. Therefore, this paper focuses on the evaluation of the classification effect of the ResNet network model from two aspects: first, the model performance is analyzed by observing the classification accuracy of the model on the JA food-1100 dataset as well as the test set in the JA photo-1200 dataset, and second, the model performance is analyzed by observing the magnitude of the classification accuracy of the model on the test dataset.

The experiments in this paper consist of 3 parts. They are evaluating the performance of the ResNet network model with the 2 different datasets A and B proposed in this paper and validating the model with the test dataset. In order to compare and see the effect of the improved models, the neural network models chosen for comparison in this paper are the classical convolutional neural network models, including VGG network and so on. All models will be trained 100, 200, 300, and 400 times.

1)

Comparison and analysis of experimental results based on dataset A

In this paper, the classical convolutional neural network and the ResNet network model proposed in this paper are first trained on dataset A. The effectiveness of the ResNet network model proposed in this paper is proved by analyzing the accuracy and loss values of different network iterations on different network models and removing the influence of the number of network iterations on the experimental results. The accuracy rate and loss rate of the model for different number of iterations are shown in Fig. 3, where (a)~(b) are the accuracy rate and loss rate, respectively. Based on the comparison of network model accuracy, the ResNet model proposed in this paper has significantly better model accuracy (0.855) than other classical convolutional neural network models when experimented on dataset A.The model loss value (0.102) of the ResNet network model proposed in this paper is significantly lower than that of other classical convolutional neural network models, which shows its superiority on dataset A.

Figure 3.

The accuracy and loss rate of the model

2)

Comparison and analysis of experimental results based on dataset B

Figure 4 shows the accuracy and loss value of the network at different number of iterations. It can be seen that when the number of iterations is 50, 100, 150, and 200, the ResNet network model has the lowest loss values of 0.778, 0.611, 0.388, and 0.177, respectively, and the corresponding accuracy rates are 0.622, 0.717, 0.736, and 0.768, respectively.Whether in terms of accuracy rates or loss values, the ResNet network model is significantly better than other classical convolutional neural network models, which all-roundly confirms the correctness and efficiency of the ResNet network on the Chu culture element dataset.

Figure 4.

The accuracy and loss value of the network when different iterations

3)

Comparison and analysis of experimental results based on test data sets

The confusion matrix of the test set is shown in Figure 5. There are 24 images in the Chu culture elements category in the test set, and 18 of them were judged correctly by the algorithm, while 6 were judged incorrectly.The total number of images in the category of non-Chu culture elements is 257, of which 232 were judged correctly and 25 were judged incorrectly.The sports culture video includes 281 images, with 250 judged correctly and 31 judged incorrectly. It is concluded that the accuracy of ResNet network model in classifying Chu culture elements in sports culture videos reached 91.43%. The extraction effect that was expected was accomplished.

Figure 5.

Test set confusion matrix

Taking a Wuhan region 18-55 years old group as the research sample of this paper, the number of its 200 people, from the cross-border cooperation, conceptual updating, cultural dissemination of the three aspects of the development of the questionnaire, in the preliminary development of the “Chu cultural elements of the integration of the effectiveness of the questionnaire” to test the credibility of the questionnaire, concluded that the questionnaire meets the standard requirements of the credibility of the questionnaire. Conditions can be used for data gathering work. Without the integration of intervention, the beginning of the formal distribution of questionnaires, the number of questionnaires issued 200, the number of questionnaires recovered 192, the recovery rate of 96.00%, excluding the invalid questionnaires 8, the effective questionnaire there are 184, corresponding to the effective rate of 92.00%. After the experimental intervention, the research sample was given the same number of questionnaires as the previous one. After the task of obtaining data was completed, the pre-intervention and post-intervention research data were saved in a point form, which facilitated the research work. In order to better highlight the effect of the integration of Chu culture elements, the method of independent sample t-test in the theory of statistical mathematics was used to analyze the differences in the integration effect of Chu culture elements before and after the intervention, and finally to verify the effect of the integration of Chu culture elements according to the significant differences in the results of the analysis.

On the basis of obtaining the research data, with the help of the independent sample t-test method in the theory of statistical mathematics, the integration effect of Chu culture elements before and after the intervention is explored, and the difference analysis of the integration effect of Chu culture elements is shown in Figure 6, where (a)~(b) represents cross-border cooperation, conceptual updating, and cultural dissemination, respectively. Based on the data performance in the figure, the difference in the mean values of the dimensions of the integration effect of Chu cultural elements (cross-border cooperation, conceptual updating, and cultural dissemination) before and after the intervention is 1.38 (3.71-2.33=1.38), 0.88 (3.07-2.19=0.88), and 1.06 (3.72-2.66=1.06), respectively, where the differences in the means before and after the intervention are most obvious in all three dimensions, followed by the differences in the mean values before and after the intervention of cultural dissemination. The most obvious difference between the mean values before and after intervention is the difference between cross-border cooperation and conceptual renewal. It can also be seen that before and after the intervention cross-border cooperation (P=0.015, T=-0.679), conceptual updating (P=0.007, T=-0.426), cultural dissemination (P=0.001, T=-2.226) P<0.05 that is, it shows that before and after the intervention of the integration of the elements of the Chu culture effect of the dimensions of the difference between the significance of the Chu cultural elements after a period of time after the experimental intervention Integration of cross-border cooperation, conceptual updating, cultural dissemination has been significantly improved, but also conducive to the promotion of sports culture and Chu culture synergistic development.

Figure 6.

Analysis of the integration effect of chu cultural elements