AI-driven Research on Lingnan Cultural Brand Innovation and Design

Lingnan traditional plastic arts inherited the traditional Chinese culture of “unity of man and nature” idea, emphasizing respect for nature, according to local conditions, thus forming a pattern of harmonious coexistence of man and nature in the Lingnan region, such as the mountains surrounded by water, the face of the screen sunrise of the Hakka villages as well as the common Guangfu-style “three tips and two corridors “layout, etc. Guided by this ideology, the traditional modeling arts in the Lingnan region are delicate and simple, pursuing natural practicality, harmony and moderation. Wood and jade carvings among traditional carvings are firstly designed from the actual situation of the materials when they are conceived and modeled, thus their modeling art is different from the uniform sculptures, and the overall form and meaning are more vivid. In addition, we can also feel the tendency of pragmatism in the pattern of Lingnan architecture, such as the Zhaolou relying on the design concept of “Western for Chinese”, the designers have integrated the architectural ideas of the Western alleyway into the design of the Zhaolou, so as to enhance the practical function of the building.

The use of color runs through the veins of Lingnan culture, whether it is sculpture, woodblock prints or costumes, none of them presents a unique visual effect through the use of color. The production of theatrical costumes incorporates the law of color contrast between warm and cool colors while adopting the delicate technique of Canton embroidery. The designers use bright and vivid colors to decorate the costumes, highlighting the colors of the costumes and distinguishing the characters through the contrast. The production of woodblock prints also applies the special overprinting process of large color blocks, such as door paintings, portrait paintings, etc., focusing on the application of red, black, blue, green and other colors, and the designer uses red as the base color to make the color combination of the paintings more bright, which in turn highlights the artistic characteristics of the paintings.

Influenced by religion, custom culture and traditional philosophical thinking, the pattern art in Lingnan culture is very delicate and contains rich humanistic connotation. According to the formal characteristics of patterns, we can classify them into figurative patterns and abstract patterns. Figurative motifs are mainly characterized by intuitive and vivid images, focusing on expressing emotions and conveying information through scenario reduction of concrete objects. Lingnan traditional architectural decorative sculpture engraved patterns of distinctive regional characteristics, carving skills, painting skills, such as engraved brick carvings “Fish Leap Dragon Gate”, “Three Suns Kaitai”, “Twin Phoenix Sunrise” and so on, from different aspects etc., which express the symbolism of the working people’s pursuit of a better life from different aspects. Abstract patterns are more focused on outlining through lines, and then draw the image language, more inclined to imagery expression, such as cloud pattern, Ruyi pattern and peony pattern.

Given that HSV is closer to real-life color perception, converting HSV to RGB can help us resolve the contradiction between the ambiguity of real life and the precise description of RGB. To wit: (1) h={ 5+b′,Pawn=max(r,g,b)And g=min(r,g,b)1−g′, Pawn r=max(r,g,b)And g≠min(r,g,b)1+r′,Pawn g=max(r,g,b)And b=min(r,g,b)3−b′ Pawn g=max(r,g,b)And b≠min(r,g,b)3+g′ Pawn b=max(r,g,b)And r=min(r,g,b)5−r′, Other situations (2) s=v−min(r,g,b)v (3) v=max(r,g,b)

Where: r′=v−rv−min(r,g,b) , g′=v−gv−min(r,g,b) , b′=v−bv−min(r,g,b) , where: r,g,b∈[0,1], h∈[0,360], s,v∈[0,1].

In the following, the three components H, S, and V are quantized in non-equal intervals according to human color perception, with hue H divided into 8 parts, and saturation S and luminance V divided into 4 parts each: (4) H={ 0ifh∈( 345,25 ]1ifh∈( 25,55 ]2ifh∈( 55,108 ]3ifh∈( 108,165 ]4ifh∈( 165,220 ]5ifh∈( 220,275 ]6ifh∈( 275,316 ]7ifh∈( 316,345 ] (5) S={ 0ifs∈( 0.1,0.65 ]1ifs∈( 0.65,1 ] (6) V=0 if v∈( 0.15,1 ]

The H, S, V intervals above are all open and closed. According to optical theory, the color of an object is related to the wavelength and frequency of light. Different color light has different ranges of wavelengths and frequencies in vacuum, so we can quantize the hue with unequal intervals. The 0,1,2,…,7 of the resultant values indicates the category of the respective hue, respectively. The three-dimensional eigenvectors H, S, V are combined into one-dimensional eigenvectors with different weights for computation [22]. Of these three vectors, the human eye’s classification of color is based primarily on hue H, followed by saturation S, and finally brightness V. The combined one-dimensional vector DD can also be taken according to the number of quantization levels of H, S, V and their bandwidths: (7) CL=HISIV+SIV+V

where I_S and I_V are the quantization levels of S and V , respectively, and taking I_S = 2, I_V = 1 according to the above method , the above equation can be expressed as: (8) CL=2H+S+V

According to Eq. (8), the value of CL ranges from an integer between [0,1,…,15], and CL is calculated to obtain a one-dimensional histogram of 16 handles. In this way, H, S, V the three components are distributed on a one-dimensional vector. The weight taken for the hue H is 8, the weight taken for the saturation S is 2, and the weight taken for the brightness V is 1, which greatly reduces the impact of the image brightness V on the retrieval results, and also reduces the impact of the saturation S on the retrieval results, but the color distribution of the image is different but it can be retrieved very well, so it can take full advantage of the color information characteristics of the image to satisfy people’s requirements for image querying. Obtain the color histogram of the image.

Histogram of the joint distribution of the HSV color feature vector CL, with the horizontal coordinate being the value of CL taken from 0,1,…,15 and the vertical coordinate being the frequency of pixels when CL occurs equal to 0,1,…,15 respectively in the whole image. From the properties of RGB color histograms it is easy to know that CL vector histograms are also rotation invariant and scale invariant.

Texture is also one of the important features of an image, and many images may show some kind of irregularity in localized areas and some regularity in the whole. This local irregularity and overall regularity in an image is generally referred to as texture. The difficulty of texture description is that there is a close relationship between it and the shape of the object, and the ever-changing object shapes and nested distributions make the classification of texture very difficult. And the co-gray birth matrix used in this study to quantify the texture features. It is described as follows:

Let p(f_k2f_l,d) denote the joint probability of the simultaneous occurrence of two pixels of gray level f_k2f_l at distance d on the image, and this probability value can be determined by calculating the number of occurrences n_kl of pairs of dots on the image at distance d, and with gray level value f_k2f₁ respectively [23]. Let n be the total number of pairs of points on the image with distance d, then the element c_kl of the covariance matrix can be expressed as: (9) ckl=p^(fk,fl,d)=nldn

Meaningful statistical features can be extracted from the covariance matrix to represent the texture. Based on this approach, energy, gray ratio, gray correlation, invertible difference moments and information entropy are found to have the greatest discriminative power.

Shape is also one of the more important features for describing an image, the contour feature of a shape we describe in terms of the perimeter of a straight line segment, and the region feature of a shape is expressed in terms of the area of the region, and in this subsection we use the Fourier descriptor to explore the shape feature.

The formula for calculating the characteristics of a straight line segment is as follows:

Number of line segments feature LN: Total number of line segments detected after Hough transform processing. Parallel line feature LP: the number of parallel lines, normalized by the total number of line segments. I.e.: (10) LP={ 0,LN=0∑i=1kNPWLN,LN≠0

Based on the special morphology, we define the region densities C to characterize them, and the region densities can be measured by the following equation: (11) C=A/P2

where P and A are the perimeter and area of the graph, respectively. According to this measure, the circle is the densest graph, which effectively reduces the error impact of the noise region on the later feature statistics. In practice, it is set according to experience: (12) QA=w*h25,Qp=1

The image connectivity domain features are calculated as follows:

Effective connectivity region number characterization: AN. Maximum area of effective region: AMA. maximum density of effective region AMC. that is: (13) AMaxC=max{ Ai }w*h (14) AMaxC=max{ C }=max{ AiPi2 }

Overall, the shape features its mainly the perimeter of the line segment and the area of the region, and the difference between the perimeter of the line segment and the area of the region will be analyzed subsequently as an illustration of the effectiveness of the shape feature extraction based on Fourier descriptors.

First the model of the image object needs to be defined. An image object I can be represented as I(D,F,R,M). Where D is the original image data. F = {f_i} is the corresponding underlying visual feature set of the image. R = {r_ij} is a collection of descriptions of a given feature (each description is itself a vector of length K : r_ij = [r_ij,…,r_ijk,…,r_ijk].where K is the dimension of this vector. In order to fully express the richness of the image, the object model allows the image to be described using multiple features (and feature representations), each of which has dynamic weights corresponding to it, and the feature fusion will be computed next using Gaussian normalization.

The visual expansion element of the brand is often closely related to the brand’s cultural connotation and brand image, which is one of the keys to increase the brand’s recognition and consumer awareness, deepen the consumers’ understanding of the brand and generate the brand’s faith. Through the above 2.3 subsection can be extracted Lingnan cultural characteristics, and take the same design style with the graphic logo, in the form of a combination of line and surface performance, and the use of color in the red, beige interlaced design application, resulting in the brand visual expansion elements of the sense of hierarchy. In order to supplement the richness of the brand image, we choose to express it in the form of illustration, dynamic lion shape and brightly colored images, highlighting the performance of dry ink texture, and the brand logo image to form a strong contrast echo.

Brand image publicity and promotion of the brand has a profound impact on the brand, in brand marketing there is an exaggerated sentence: one point to do, nine points of publicity, very perfect. Good brand image and the right way to promote the brand can play a role in expanding the consumer market, the importance of the brand’s image is self-evident. In addition to the brand itself, based on the development of regional cultural development of the brand in the development process can also effectively enhance, show its regional image, and pull the regional tourism economic growth.

Cultural and creative derivatives are also the expression of corporate brand image through the medium of articles, which is another means of publicity and promotion for the cultural connotation and value of the corporate brand. By combining and applying the colors, textures and shapes of Lingnan culture to various office supplies, clothing and brand peripherals, the brand is invisibly disseminated during the process of using them, which becomes a mobile publicity medium, and at the same time enhances brand image and public recognition of the corporate brand. At the same time, it can also enhance the brand image and the public’s sense of identification with the enterprise brand, expand the consumer group and promote consumption. Lingnan culture brand of cultural and creative derivatives of the basic combination of color, texture, shape as the main design elements, the overall color tone of red and beige, with auxiliary color design, to create a bright and harmonious design sense, from the visual touch consumers, let a person memorable.

The number of colors most commonly used in Lingnan cultural and creative designs is 7~8, including red, orange, purple, blue, green, green, yellow, white, brown, rosy red and other colors, in order to be able to more comprehensively express the richness of the color of Lingnan cultural and creative patterns, and taking into account the need for color matching choices in the design, the color information of the Lingnan cultural and creative designs is set at 16 according to the maximum number of colors of a single piece of twice the number of maximum value. Using the HSV color space model, the color space model is used to describe color characteristics from the perspectives of hue, saturation and brightness. Using the HSV color space model to describe the color characteristics from the perspective of hue, saturation and lightness, the color matching and composition law of Lingnan cultural and creative design is visualized with quantifiable data, and the color information of Lingnan cultural and creative design is shown in Table 1. According to Table 1, the proportion of warm colors, mainly red, yellow and orange, is about 73%, and the proportion of cold colors, mainly blue, purple and green, is about 27%. The brightness and saturation of 1, 2, 4, 5, 6, 8, 9, 10, 11, 15, and 16 are in the medium-high dimension. 13 and 14 are low brightness and low saturation colors. 3, 7, and 12 are low saturation and high brightness colors. It can be preliminarily inferred that in Lingnan cultural and creative patterns, warm colors with medium-high brightness and saturation are used more and have a larger area in the picture, followed by cool colors. The pastel colors with low saturation and high brightness and the dark colors with low brightness and low saturation are used in a smaller area, and are mostly used as accent colors in the whole picture.

In order to better explore the color matching relationship and color usage frequency of Lingnan cultural and creative patterns, this paper constructs a color network model with primary color priority to visually reflect the color matching law of primary and secondary colors in Lingnan cultural and creative. A node on the color network represents an extracted color, and the larger the node indicates that the color appears more frequently in all colors. The connecting line between colors appears only when the threshold is exceeded, and the connecting line indicates that the connected colors co-occur in the same picture at a certain frequency many times to produce the collocation relationship. The construction of the color network model in this paper is done using a programming language, and it is plotted in the Matplotlib plotting library in Python, and the co-occurrence threshold is set to 10%. The color network model of Lingnan cultural and creative patterns is shown in Figure 1, and the three color matching schemes with the No. 01 color as the main color are shown in Figure 2. The rich and bold use of colors in Lingnan cultural and creative designs is also evident from the connecting lines of the collocation relationships in the color network model. Among them, the connecting lines between the four colors 01, 03, 05 and 13 are significantly thicker than the connecting lines between the other colors, indicating that the frequency of co-occurrence is high, the color relationship is close, and it can represent a certain number of color matching styles in the Lingnan cultural and creative patterns, and the designers can focus on considering the above colors as the main color matching scheme in the color selection. Designers in the actual design process often need to determine the number and proportion of color matching according to the demand, so this article to 01 color as the main color and with 2 kinds of secondary colors as an example, will be all the matching scheme to show in detail, with the increase in the number of colors, with the choice of options will be more abundant, color mapping can provide designers with hundreds of different color schemes, greatly enriching the designer’s color scheme options! The color map can provide designers with hundreds of different color schemes, greatly enriching the designers’ color matching options.

Figure 1.

Lingnan cultural and creative pattern color network model

Figure 2.

Three color matching schemes with color 01 as the main color

In this experiment, three different textures of Lingnan cultural and creative images (A, B, C, each with three images, a total of nine images) are selected, converted to 512 pixels × 512 pixels BMP images by software, and the texture features are calculated using the symbiotic matrix, and the results of the texture feature calculation are shown in Table 2. From the parameter energy analysis, the energy is the sum of squares of the values of the elements of the grayscale symbology matrix, the wall tile texture in the generic texture library and the stone texture in the self-constructed texture library are selected for comparison. The energy values calculated for the three images in the A texture are very equal, with a maximum difference of 0.013, and the energy values calculated for the three images in the B texture have a relatively large difference, with a maximum difference of 0.062 and the energy values of the B texture are all higher than the energy values of the A texture. Finally the maximum difference in the energy value of C texture is 0.011 and its maximum difference in energy is lower than A and B textures respectively. From the experimental images it is concluded that the texture represented by the three images of stone texture is not as regular and homogeneous as the wall tile texture. This is because the energy reflects the uniformity of gray scale distribution and texture coarseness of the image. If all the values of the covariance matrix are equal, the energy value is small. Conversely, if some of the values are large and others are small, the energy value is large. When the elements of the symbiotic matrix are centrally distributed, then the energy value is large, and a large energy value indicates a more homogeneous and regularly varying texture pattern, numerically demonstrating three different texture characteristics. Due to limited space, the other four parameters are not repeated. By calculating the five parameter values of entropy, energy, inverse disparity, gray scale correlation and gray scale ratio, it reflects the effect of the application of symbiotic matrix in texture features.

This subsection selects 20 Lingnan cultural and creative images with different shapes as the object of this research, based on subsection 2.3.3, it can be seen that the shape feature investigation indexes are line segment perimeter and area, the calculation results of this paper are regarded as the theoretical value, while the test results of the professional equipment are the real value, and the comparative analysis of the calculation results of the shape feature is done to detect the effectiveness of the Fourier descriptor in the shape feature extraction, and the results of the calculation of shape features are The comparison is shown in Figure 3, where (a)~(b) are the perimeter of the line segment and the area of the region, respectively. It can be seen that the difference between the real value and the theoretical value is controlled within 5%, i.e., the Fourier descriptor has obvious application effect on the shape feature extraction, whether it is the perimeter of the line segment or the area of the region.

Figure 3.

Comparison of calculation results of shape characteristics