Extraction and Protection of Historical Architectural Features of Traditional Villages Based on Computer Vision

The geometric principle of using UAV tilt photography to collect the basic data of historical buildings and cultural relics is mainly based on the covariance equation, using the beam method of leveling, with the help of the internal orientation elements, solving the coordinates of each point in the image at the moment of aerial photography by UAV, realizing the relative positioning of the common point between the pixels, and then converting the image coordinates into spatial coordinates by coordinate transformation.

When the center of the photo is on the same line with point p and point P on the ground, the expression is shown in (1): 1x−x0=−fm1(XA−Xs)+n1(YA−Ys)+p1(ZA−Zs)m3(XA−Xs)+n3(YA−Ys)+p3(ZA−Zs)y−y0=−fm2(XA−Xs)+n2(YA−Ys)+p2(ZA−Zs)m3(XA−Xs)+n3(YA−Ys)+p3(ZA−Zs)}

Where: x₀, y₀, f denote the internal parameters of the camera, X_A, Y_A, Z_A denote the coordinates of the point A that has been determined in the ground coordinate system, X_s, Y_s, Z_s are the coordinates of the imaging center of the camera in the ground coordinate system, m, n, p are the direction cosine values of the azimuthal unit, and x, y denote the points to be measured in the ground coordinates. The key consideration to focus on is that the UAV has to surround the building to control the flight during the flight, and if the number of buildings is too many and the scale is too large, then the flight height of the UAV has to be planned to ensure that it can complete the photography of all the individual building bodies. The sailing height of the UAV can be calculated according to equation (2): 2H=h×GSDq

Where: H denotes the navigational height of the camera relative to the ground (m), h denotes the focal length of the camera (cm), q denotes the pixel size, and GSD is the ground resolution of the image.

After completing the data collection of historical buildings and cultural relics, its data is pre-processed, which is divided into the following steps. 1)

Cropping: manual cropping is carried out first, which can not only reduce the amount of data, but also facilitate the processing. After completing the cutting process, the data will be output, and the file formed is applicable to TXT format and DXF format.

For the characteristics of the building plane features more, according to the distance from the point to the plane to determine whether to eliminate the point. Let the space plane equation be expressed as: 3ax+by+cz=d

Where: (a,b,c) represents the unit normal vector of this plane, d represents the distance from the origin of coordinates to this plane, d ≥ 0. After completing the above operations, the data are then subjected to a roughness clearing operation as follows. 1)

Calculate the initial optimal plane by combining the initial values of the three parameters a, b and c of the plane equation.

Through the reverse engineering software, the article first establishes the triangular mesh surface of the processed data according to the nearest principle to form a prototype close to the physical building, and then with the help of the surface fitting method [20], the 3D model is directly formed through the processing of repairing and smoothing. The historical building’s 3D model is created using 3D reconstruction technology based on the processed data. In this paper, the specific approach for modeling historical architectural artifacts is to start from data and construct surfaces using accurate mathematical functions. Then the interpolation method and approximation method are used to interpolate other unknown data points according to certain principles or functions based on known data points, and the approximation method is used to construct the surface as follows: 5f(x,y,z)=a1x2+a2y2+a3z2+a4xy+a5xz+a6yz+a7x+a8y+a9z=0

When modeling surfaces, the surface is first fitted to the measured data, and then the resulting surfaces are transitioned, blended, and connected to form the final surface model.

After completing the 3D model needs to be finely mapped according to the model contour lines, when the mesh model is represented as a mesh, texture mapping technique is used to determine the corresponding positions (u, v) of the scene surface P and the actual points P [21], while the texture or color defined on (u, v) is used to portray a certain texture feature of the scene object at the P point. The texture mapping technique can be divided into two steps. (1) Determine which parameters of the surface need to be defined as different texture features. (2) On this basis, map the texture space to the scene space and screen space.

The advantage of fine mapping images of historical architectural relics is that it can provide accurate three-dimensional coordinate data and texture information, which provides more reliable data support for the protection and research of historical architectural relics.

In this paper, HSV color space model is selected for extraction of color features. Generally, images are saved in RGB, while in this paper, images are converted from RGB color space to HSV space. Let V_c is a color in RGB color space and W_c is a color in HSV color space, the transformation process from RGB color space to HSV color space can be represented by equation (17) [24]: 17Wc=Tc(Vc)

Where T_c represents a linear transformation, which is the transformation process that converts the color from RBG space to HSV space. If r, g, and b in the RGB color model are normalized, the values of h, s, and v of the HSV model after the linear transformation will also be in the [0,1] range. The T_c transformation is represented as follows: 18h={ 5+b′, When r=max(r,g,b)And g=max(r,g,b)1−g, When r=max(r,g,b)And g≠max(r,g,b)1+r, When g=max(r,g,b)And b=min(r,g,b)3−b, When g=max(r,g,b)And b≠min(r,g,b)3+g, When b=max(r,g,b)And r=min(r,g,b)5−r, Other casesv=max(r,g,b)s=v−min(r,g,b)v

Among them: 19r1=v−rv−min(r,g,b)g1=v−gv−min(r,g,b)b1=v−bv−min(r,g,b)

Color histograms are often used to represent the color features of an image with fast computation, simple algorithms, and scale, translation, and rotation invariance. [25]. This method has good applications in image feature extraction, image classification, and retrieval. The algorithm is able to represent the distribution of different colors and the proportion of colors in the whole image, i.e., the colors contained in the image and the probability of occurrence of each color.

The color histogram of an image can be viewed as a one-dimensional discrete function with the following formula: 20H(k)=nkNk=0,1,…,L−1

Where k represents the color grade of the image, L is the number of desirable color grades. n_k is the number of pixels in the image having color level k, and N represents the total number of pixels in the image.

The Chating-Lianjiang Waterway Ancient Road is taken as an example to be analyzed. According to the results of the correlation analysis of the elements of the zongshi forms along the Chating-Lianjiang Waterway Ancient Road, there are 9 pairs of significant correlations in the zongshi forms along the Chating-Lianjiang Waterway Ancient Road, and the correlation analysis table of the elements of the zongshi forms along the Ancient Road is shown in Table 2.

In the sample, 13.9% of the zongshi axes were turned, and all of them were oriented to the north. To further explore the reasons from the research subjects, there are two cases of the axial turning of ancestral halls located in the northern part of the Xingzi Ancient Road, which are the Tang Clan Ancestral Hall in Lou Village and the Yi Clan Ancestral Hall in Yi Jia Village, which are closely connected with the residential buildings, and are large in size, and it is difficult to leave a square for activities in front of the head gate, and so they are surrounded by a wall to form a courtyard, which is a common practice in part of the region of Jiangxi Province. The axes of 86.6% of the ancestral halls are still in place, with 48.5% facing southwest and only a few facing east or north.

Along the ancient road, 72.78% of the samples have patios and 13.38% have corridors, and most of the samples with corridors are located in Hakka huts, which appear in the southern part of the ancient road. The architectural beauty of this type of ancestral halls is insufficient, and a large number of examples are not included in the list of immovable heritage, but the author believes that this type of ancestral halls with corridors can be regarded as a typical form in the southern part of the Chatting-Lianjiang Waterway Ancient Road. Among the ancestral halls with patios, raised beam structures are widespread, accounting for 87.67% of the total, and 87.38% of these raised beam buildings with patios do not have moon pools, 6.44% have half-moon moon pools, and 6.21% have other types of moon pools. The Chaoyang He Clan Ancestral Hall in Qingjiang Village, which has a half-moon moon pool, was built in the early Qing Dynasty and rebuilt in the Qianlong Dynasty. The overall shape conforms to the common features along the northern part of the ancient road mentioned above, and the presence of moon pools can be seen as one of the strong evidences of the integration of Hakka elements. Among the samples with corridors, 66.32% have half-moon moon pools, and 33.84% have other moon pools.

Along the ancient road, 95.1% are integrated ancestral halls, and in the research case, only the Luo’s Yuzhang Hall in Shuiyuan Village has the phenomenon that the head gate is disconnected from the main body of the building. Lianzhou villages often have a gatehouse at the entrance, and from this point of view, the Luo ancestral hall can also be regarded as a gatehouse in front of the ancestral hall and a monolithic building. However, most of the gate towers in Lianzhou have eight-character walls, and there is a half-moon moon pool in front of the gate tower of Luo’s ancestral hall, so the author discusses this case as one of the few cases in which the head gate and the building are separated in Lianzhou. 95.3% of the remaining ancestral hall buildings are three-roomed buildings, and the outer boundaries of the buildings are flat with no protrusions. From the above relevant analysis, we can depict the general characteristics of the ancestral halls along the Chating-Lianjiang waterway: there is a square in front of the ancestral hall buildings along the ancient road in the northern Lianzhou and Yangshan territories, and there is no moon pool on the square. The face is three rooms wide, with a lifting beam structure. The outer boundary of the building is level. In the south, along the ancient road in the territory of Yingde, there gradually appeared the horizontal hall house type residence and rituals in one building form, this kind of ancestral hall building and the same material used in ordinary residential houses, is located in the center of the horizontal hall house, the area is small, the head gate, the middle hall and the sacrificial hall are separated by the horizontal hall house corridor, single room or three rooms are available, the area is small. The outer boundary of the building is connected to the residence, which shows concave and convex changes along with the functional facade of the residence. In front of the plaza, there is a half-moon-shaped moon pond. The two types are not limited to a specific region, as there is a fusion of architectural forms along the ancient road.

The existing traditional houses in this village are mainly built with traditional building materials, and the chosen materials are mainly red bricks, granite stones, and red tiles. The colors of these traditional materials themselves make the traditional houses form a unique color style. The color analysis of the traditional residential buildings, the analysis results are shown in Figure 3.

Figure 3.

Distribution of traditional residential houses color composition

The roofs and red tiles are mainly in the range of R (red) and YR (yellow-red) hues, which are warm and can bring a warm feeling, and the walls are mainly N (no color), Y (yellow) and GY (yellow-green), which are inclined to warm tones, and the overall landscape has fewer kinds of colors, with a simpler match. The brightness and color values of all the colors are in the middle and low grades, giving people a subtle and calm feeling.

The village is a traditional village with comprehensive beliefs in Buddhism, Taoism, and folk beliefs, etc. There are more than ten temples in the village, as well as a number of ancestral halls and ancestral house buildings with different family names. Most of the temple buildings with color is relatively bold, gorgeous and rich, with white stone wainscoting, wall plugs, counter feet, etc., with red bricks into the body of the building walls, with white or green stone carved into the bar Hoven windows, chi dragon windows, etc., the roof with brick-red tiles, the roof, the walls, porches and other places there are a number of painted decorations, colorful and gorgeous, the overall building in red is the most eye-catching. Ancestral buildings are similar to traditional residential buildings, compared to residential buildings, they are larger in volume and more opulent in color, focusing on the ridge decoration and stone and brick carvings on the walls, with red and off-white as the main colors.

Quantitative analysis of religious building colors, the results are shown in Figure 4. Red bricks and red tiles are mainly R (red phase, a warm color system, the temple’s green stone carving columns are B (blue), BG (purple-blue phase, a cold color system, the contrast between cold and warm, especially eye-catching. White stone carvings, granite oyster shells are mainly N (no color), G (yellow-green), Y (yellow) color phase, architectural color decoration is very rich in color. Brightness, basically belongs to the range of medium brightness, R (red) color value is concentrated in the medium color level, the rest of the color phase are low color ridge of the decorative color color and brightness is higher to play a certain role in the embellishment.

Figure 4.

Distribution of traditional residential houses color composition