The Integration Path of Innovative Design Thinking and Rural Development of Non-Heritage Cultural Creations in the Era of Artificial Intelligence

There are three main concepts of design thinking. The first view is that design thinking is a means of problem-solving, and that design thinking can creatively solve difficult problems that are characterized by ambiguity and diversity of solutions. The second view is that design thinking is a method of innovative thinking. Design thinking is an iterative innovation process that focuses on user needs, including empathy, definition, prototyping, and testing. The third view is that design thinking is a skill specific to designers, centered around their way of thinking, which is centered on the user and related to the nature of design work. By comparing the three perspectives, this study concludes that design thinking is a collection of thinking modes aimed at solving design problems, in the spirit of user-centeredness and innovation-driven main lines.

The general law of the development of things is from nothing to something, from fragmentation to stability, and so is design thinking. The generation mechanism of design thinking is shown in Figure 1, which is divided into three stages: occurrence, formation, and perfection, and constitutes the whole process of design thinking generation with three important links: experience-empathy, interaction-integration, and detection-iteration. In the occurrence stage, design thinking starts from the empathy perspective for user experience, and the output is the direction. In the formation stage, design thinking is based on team thinking and visualization tools, and the outcome is creativity. In the refinement stage, from testing to iterative optimization, the output is the solution. In the occurrence mechanism of design thinking, design thinking starts from empathy, emphasizes user-centeredness, grasps user needs through user observation and in-depth interviews, defines the problem, and produces the direction of design.

Figure 1.

Design thinking generation mechanism

The ultimate of non-heritage cultural and creative products is to face the general consumer group, so the design of non-heritage cultural and creative products also needs to combine technical thinking and commercial thinking on the basis of design thinking, and comprehensively consider the realistic factors for creation and development.

In design thinking, the experiential thinking method refers to starting from the user’s real experience, digging deep and satisfying the user’s emotional demands, and delivering an interactive, cognitive, and emotional personalized experience. This experience-centered design thinking not only requires designers to establish a close connection between designers and users to achieve a deeper involvement with each other, but also requires designers to start from the users’ point of view and pay attention to their real feelings and needs. Identifying and satisfying needs is the core of experiential thinking, where needs include emotional, behavioral, and functional needs of products mentioned in the user-thinking approach. Therefore, the experiential thinking approach in design thinking is a design thinking approach that focuses on users’ real experiences and needs. It can not only improve the effectiveness and quality of design, but also encourage the transformation of design process management, so that the design is more aligned with the needs and expectations of users.

Business thinking is premised on assumptions, based on facts, and verified through logical analysis. The corresponding thinking methods include the logical thinking method, which emphasizes analytical thinking, and the decision-making method, which emphasizes crisis awareness. According to the theory of dual processing of information, human decision-making thinking can be divided into an intuitive system and an analytical system. These two systems have different characteristics and ways of processing information and making decisions, and business thinking is more inclined to analytical thinking. In business thinking, analytical thinking can help decision makers analyze problems more rationally and objectively, assess risks and opportunities, and formulate more reasonable decisions. Judgments are formed through in-depth analysis of problems, collection of relevant evidence, and logical reasoning. This way of thinking is systematic, organized, and precise, and can ensure that decision makers maintain a clear mind and an objective attitude when dealing with complex business issues. The advantage of analytical thinking is that it helps decision makers rationally assess risks and opportunities, and avoid blindly following trends and making impulsive decisions. Through logical reasoning and data analysis, decision makers can more accurately grasp market trends and anticipate future changes, and thus formulate more reasonable and feasible business strategies.

Based on the background of the artificial intelligence era, this paper proposes an innovative design thinking mode that synthesizes technical thinking and business thinking, and the innovative design thinking mode of non-heritage cultural and creative products is shown in Figure 2. The innovative design thinking of non-heritage cultural and creative products has a strong synthesis, which integrates technical thinking and business thinking to make the non-heritage cultural and creative products innovative, technical and at the same time in line with the market demand and business objectives.

Figure 2.

Innovative design thinking model of non-relic products

In the context of the artificial intelligence era, “creative transformation and innovative development” is the guideline for the protection and inheritance of non-legacy, and the use of modern science and technology to realize the digital protection and development of non-legacy is one of the most important ways for the innovation and inheritance of “non-legacy”. Introducing artificial intelligence technology into the workflow of non-heritage cultural creations to improve the design efficiency of non-heritage cultural creations, increase the communication contact and promote the commercial transformation is the core concern of this research. Artificial intelligence can achieve a certain level of automation and generate multi-creative solutions with high efficiency. At present, the research on the contact and transformation of non-heritage cultural and creative products mainly focuses on extracting elements. From the perspective of element extraction, the images and color elements of non-heritage are directly quoted, simplified and extracted, and deconstructed and reorganized to be applied in different types of cultural and creative products, from the origin of the craft to the inheritance of the technique to the expression of the final cultural value, and step by step in-depth to complete the design of the cultural and creative products, and the process of introducing the non-heritage cultural and creative process assisted by the artificial intelligence technology is shown in Figure 3. After the material of non-heritage cultural creation is taken, the algorithm model is built with the assistance of artificial intelligence technology to identify and extract the design elements such as action, image, color, etc., and then carry out the design of non-heritage cultural and creative products through the deconstruction and reorganization of the design elements, which saves the designers’ time of collecting and extracting the design elements in the early stage, and at the same time improves the design thinking of the cultural and creative products through the design of the elements compared to the screening, greatly improves the design of the cultural and creative products, and improves the design of the cultural and creative products. It greatly improves the efficiency of non-heritage cultural creation.

Figure 3.

The introduction of artificial intelligence technology aided non-legacy process

The process of non-heritage cultural and creative products involves the use of materials, images, and colors of cultural and creative products, etc. Different types of raw material selection, molding process, decorative process, etc. create different types of non-heritage image characteristics, and the application of artificial intelligence technology in the extraction of cultural and creative design elements can effectively improve the efficiency of cultural and creative product design. Artificial Intelligence (AI) is a complex technology and methodology designed to simulate human intelligence, including perception, reasoning, learning, understanding, and decision-making. Deep learning is a machine learning method in the field of artificial intelligence that learns feature representations through multi-layer neural networks to achieve automatic classification, recognition, segmentation, and other tasks on data. Convolutional Neural Network (CNN) is recognized as the main architecture for deep learning because it discovers its unique features by performing a learning process on the raw data, CNN shows high performance in problems such as classification, recognition and segmentation, due to this reason, many of its domains are becoming widespread and are becoming more popular in the use of big data especially due to the automatic feature discovery.

In this study, a new network unit with grouped convolution and channel shuffling is designed for NRM, which significantly reduces the number of parameters in the model and achieves good performance. Secondly, a CBAM attention mechanism is added to the network unit, which can extract key features by weighting the output features in space and channel.

Image recognition classification is a fundamental technique in the field of computer vision. It aims to minimize classification errors by distinguishing different types of images based on their semantic information. Traditional image classification methods mainly classify images based on their features, this method is essentially feature based learning for image classification, the most important step is feature extraction, in this phase specific parts of the image are encoded using manually designed algorithms such as shapes, colors, and materials, image classification is usually faced with a challenge in three areas, complex image backgrounds, identical shape classes, and Low quality images, these features are used to evaluate the image content. Deep learning based image recognition classification task relies on the powerful learning capabilities of neural networks to automatically learn and extract relevant features of an image, by training a large amount of data, the model learns the rules of the features and then predicts the data categories based on different criteria, if the prediction is incorrect, it will be corrected in time. During successive iterations, the optimal effect is searched for and then the network performance is evaluated by a validation set. The deep learning-based approach mainly extracts desired features through convolution and determines classification categories through a classifier. The specific formula is as follows: 1xjl=f(∑i∈MiXil−1×Kijl+Bt) 2xjl→Softmax→pi=(pi,1,pi,2,pi,3,⋯pi,i⋯pi,n) 3∑ pi,j=1

Conventional convolution is a way of channel-dense concatenation, where convolution operations are performed together on each channel of the input features, and then, the size of the convolution kernel is specified as K×K, the number of channels per convolution kernel is denoted as C, the number of convolution kernels is defined as N, and finally, the number of channels for the output features is also set to N, and the expression for the parametric quantity P is given by Eq: 4P=K×K×C×N

The aim is to change the input feature channels from group C to group G, the convolution kernel is also divided into group G, the number of input feature channels is changed to C/G, and after the grouped convolution operation, all the channels are spliced together to obtain a large feature map with N channels, and the expression for the number of covariates is as follows: 5PGC=K×K×K/G×N

In this study, a function classification model is proposed which consists of three components labeling smoothing loss function, attention mechanism module and LeakyReLu activation function.

The Leaky ReLu function is chosen as the activation function for the study. Although the ReLu function is widely used in advanced deep learning networks, the drawbacks still exist, especially the ReLu activation function tends to lead to a function value of 0 on one side, which is directly equal to 0 when the gradient value is negative. The equation of the ReLu activation function is as follows: 6Relu(x)=max0,x={ xifx>00otherwise

Therefore, Leaky ReLu function is chosen as the activation function with good performance to alleviate the above problem with the following equation: 7LeakyReLu(0,x)={ xifx>0Axifx≤0

In this study, Perry’s Handbook images of various simple and complex functions, with the simplest category being something like multicurve functions, which is a relatively large percentage, and the most complex category being the liquid-phase plots, which is a smaller but indeed complex percentage, with a serious imbalance in the data categories. Therefore, in this work, the labeled smoothing function is used as a loss function, and the CE-based loss function is constructed as: 8LossCE=1N∑n=1NlogPnj where N denotes the number of samples and P_nj is the probability that the N th sample is of class i. The function of label smoothing loss is defined as: 9LossLS=1−εCE(i)+ε∑ ce(j)N

In order to calibrate the recognition effect of the intelligent recognition model based on deep learning in this paper, the following evaluation indexes are introduced to assess the recognition effect of the trained deep learning model. Accuracy (Acc) is the percentage of the total samples that are predicted to be correct results, which allows for the assessment of overall correctness. Precision means the percentage of all samples predicted to be positive that actually contain samples that are actually positive. Recall is also known as the percentage of samples that are actually positive, i.e., the percentage of samples that are predicted to be positive. F1 value is a comprehensive evaluation index that takes both precision and recall into account, so that both precision and recall are as high as possible, and a relative balance is achieved.

In this paper, deep learning modeling training is carried out on the collected non-legacy image dataset, and the change in accuracy and loss value of model recognition is shown in Fig. 4. After the model starts to be trained, the accuracy rate of the model rises considerably, and then there is a small downward trend in 35 rounds of iteration, which is because when the model is trained to seek for the global optimal solution, it needs to skip some local optimal solutions, to avoid the program to have an underfitting state This is because the model needs to skip some local optimal solutions when it is trained to seek the global optimal solution to avoid the underfitting state of the program, and this process may lead to a temporary decrease in accuracy. The recognition model tends to stabilize after about 40 rounds of iterations, at this time, the model training needs to be stopped in time to save the optimal recognition model, and the overall recognition accuracy of the model is 97.35%. The loss value of the model shows an overall decreasing trend which indicates that the error between the predicted value and the true value of the images imported in each batch of the convolutional neural network is decreasing. After 35 rounds of iterations, the loss value of the model stabilizes and the loss value is below 0.5. When the error value becomes large, the internal parameters of the convolutional neural network jump in the opposite direction to keep approaching the optimal model.

Figure 4.

The accuracy of model recognition and the change of loss value

The model of this paper is used to recognize and extract different non-heritage images, including ornaments, totems, flowers, birds, people, and costumes, etc. The recognition and extraction effects of the intelligent model of this paper are shown in Table 1. The accuracy rate of image classification for different non-heritage categories is 96.98%, 90.63%, 91.92%, 92.32%, and 93.43%, respectively.

From the experimental results, it can be found that the recognition effect of totem and bird images is worse, while the recognition and extraction effect of ornament images is better, with the highest F1 score of 0.961. Taken as a whole, the average accuracy rate of this paper’s intelligent model in the recognition and extraction of non-heritage images is 93.056%, with an F1 of 0.933, which is excellent in the overall recognition effect, and it meets the actual demand of the extraction of non-heritage cultural and creative design elements.

In order to explore the practical application effect of this paper’s innovative design thinking on non-heritage cultural creations with the assistance of artificial intelligence, this study takes the consumer’s point of view as the evaluation scale to study the feasibility of the application of AI-assisted non-heritage cultural creation design.

In the previous paper, the way of application of today’s AI technology in non-heritage cultural creations has been explored, and an intelligent identification model has been constructed to study the organic combination of intelligent technology and the design process of non-heritage cultural creations. Next, this paper will explore the impact of AI technology tools on non-heritage cultural and creative design practices in greater detail. The research was conducted using a questionnaire that was distributed through a network for data collection, and a total of 148 valid questionnaires were distributed and recovered. In order to more accurately and intuitively analyze the degree of consumer acceptance of AI-assisted non-heritage cultural and creative design products, a five-point scale was used, with a minimum score of 1 and a maximum score of 5. Among them, the higher score of the scale represents the higher acceptance, i.e., the higher recognition of the design of the option. In this paper, a total of five groups of non-heritage cultural and creative products are provided according to the proposed non-heritage cultural and creative design method and are noted as AIRT1~5. The results of the satisfaction evaluation of the non-heritage cultural and creative products are shown in Table 2, and the satisfaction scores of the five groups of non-heritage cultural and creative products are, in descending order, AIRT2 (3.76), AIRT3 (3.52), AIRT5 (3.51), AIRT4 (3.45) and AIRT1 (3.07). It can be found that there are some differences between the five groups of cultural and creative products, but the satisfaction scores of the cultural and creative products in each group are all above 3 points, and the comprehensive evaluation score of the five groups of products is 3.462, which indicates that all of the non-heritage cultural and creative product designs based on the AI method in this paper have good practical application effects.

In order to further study the difference between the design of non-heritage cultural and creative products and the traditional design method under the assistance of artificial intelligence, this paper conducts a comparative experiment for the image design in non-heritage cultural and creative, designing three groups of non-heritage cultural and creative products according to the intelligent assisted design method in this paper, numbered as the intelligent group AI01~AI03. and then letting the same designers carry out the design of three groups of non-heritage cultural and creative products by traditional means, numbered as the sample group CN01~CN03, and the comparison of the satisfaction of different non-heritage cultural and creative products is shown in Table 3. CN03, the comparison of the satisfaction of different non-heritage cultural and creative products is shown in Table 3. It can be seen that the scores of the three groups of cultural and creative product designs of the intelligent group are higher than those of the sample group, which are 3.48, 4.12 and 4.05, respectively, and the comprehensive average score of the intelligent group is 3.883 compared with that of the sample group, which is an increase of 0.657. Therefore, it can be proved that the innovative design of the non-heritage cultural and creative products with the assistance of AI in this paper is more popular among the consumers, and it has a higher degree of practicability.