Abstract
Color is an important index for human visual evaluation of landscape, and it is also a key factor affecting people’s recognition and experience of heritage landscape. In this study, five important sites of the Yangzhou Grand Canal were selected for the color quantification analysis by using the Deep Learning(DL) scene parsing algorithm. The color characteristics of the winter and summer landscape of the five sites were evaluated as well as the Scenic Beauty Estimation (SBE) value. Furthermore, the correlation analysis between the color characteristics and the SBE value was established in order to study the relationship between color characteristics and the landscape beauty. The main results are as follows: ①.The dominant color of the five sites is blue and green, the building color is mainly orange and yellow in both winter and summer. The dominant plant color in five sites is green in summer, whereas in winter, changes to yellow(Site5:YZJGD) or cyan(Site1:DGGD, Site3:GZGD); ②.The overall color saturation is low in winter with the percentages of Very Low Saturation in almost each site(except site5:YZJGD)reach 80–98%. Summer has Medium Saturation colors, the percentage of Mid Saturation of sky in Site 2(GMS) in summer is 44.87%. ③. The landscapes have low brightness in winter and higher brightness in summer in all sites, sky is the only category whose High Brightness value exceeds 50% in both seasons.And in winter, landscapes are most prevalent in Low Brightness and Medium Brightness. In summer, the percentages of Medium Brightness and High Brightness increase.④.The color diversity of the sites in winter varies significantly, whereas the color diversity of the sites in summer varies slightly.The highest color diversity of plants is found in DGGD(Diversity > 1.5). ⑤.In winter, the highest SBE value is found in Site2:GMS(0.5956), and the lowest SBE value is found in Site5:YZJGD(− 0.8216),which is a large gap(1.4172).The highest average SBE value is in Site2:GMS(0.5062), followed by Site3:GZGD (0.2091), which both have average values greater than zero. ⑥.Correlation analysis revealed that there is no significant correlation between the saturation and SBE values(p > 0.05).However, the Pearson correlation coefficients which are − 0.625(winter) and 0.689(summer) indicate strong correlation.Meanwhile, there is no significant correlation between the color diversity and SBE values(p > 0.05). However, the Pearson correlation coefficients are 0.807(winter) and − 0.747(summer), indicating strong correlation.This study provides an in-depth examination of the Canal landscape color, it is hoped to promote the systematic and scientific study of landscape colors and provide a theoretical basis for the scientific design of heritage landscape color.
Similar content being viewed by others
Introduction
The Yangzhou Ancient Canal, which is the oldest section of the Grand Canal in China, has been listed as a World Heritage River by the United Nations Educational, Scientific, and Cultural Organization and has received significant attention from interested parties. Most related research on the Grand Canal focuses on the historical study, ecological protection, planning, and management of the canal, among which research on the visual perception of the Grand Canal is seldom reported. Over time, the function of the Grand Canal has changed from a shipping function to heritage and landscape functions. The heritage landscape provides a special visual experience associated with strong historical significance and cultural emotions. Canal landscape color is an important factor affecting the visual perception and experience of the canal, which reflects the cultural history of the canal and affects its aesthetics. The landscape color of the canal is determined by the canal itself as well as heritage buildings and plants, which convey the special humanistic spirit and artistic aesthetics of the Grand Canal.
Relevant research has shown that the most important sense of the human body is visual perception1 and that vision is more direct and effective than hearing, smell, touch, or other sensations1,2. Color is the first element of human perception and is more likely to attract attention and provide rich visual stimuli2,3,4,5. Color gives an instantaneous and long-lasting impression as color is the first type of information to stimulate the visual nerve6.
The famous color psychologist Angella Wright indicated that, given the right amount of information, patterns for combining different colors in a scientific manner are computable7,8. The collection and recording of colors is the basis for the quantitative analysis of landscape colors. Traditionally, landscape colors have been collected using colorimetric cards, color meters, or digital cameras9,10,11,12,13. The colorimetric card method, which is more subjective, is suitable for collecting and processing a large amount of color data. Color meters, which face certain instrumental limitations, are only suitable for measuring the color of a landscape up close. Digital cameras take photos, which can then be processed using image-processing software (e.g., Photoshop, ColorImpact, and Matlab) and are suitable for the collection of color over a large area4,5. For the quantification of color, the common color models are the RGB (red, green, blue), CMYK (cyan, magenta, yellow, black), CIELab (CIELab color model), and HSB (hue, saturation, brightness) models. Each model corresponds to a different color medium or applicable object, and different color models lead to some variability in landscape color evaluation results14. Based on the different color models, there are two main categories of landscape color evaluation. One is based on the various primary colors expressed by materials and uses Photoshop, ColorImpact, and other software to extract RGB values for color analysis14. Additionally, the CIELab color space is used to conduct quantitative studies of landscape color. The other category is the table color system, which is based on the discriminative ability of the human eye to reflect complex and diverse combinations of landscape colors15. Most research focuses on the use of apochromatic systems to classify landscape colors. However, no uniform standard for the classification basis or selection of color systems exist, and the evaluation of landscape colors using different types of color systems leads to differences in evaluation results14,16.
With the development of artificial intelligence (AI) and image processing technology, there have been new breakthroughs in the quantitative research of color. Research on color quantification using AI, machine graphics, and image technology has been conducted in the fields of architecture, botany, advertisement design, art, etc15,17,18,19,20,21,22,23,24. Computer vision technology involves the use of cameras and computers to replace the human eye for target recognition, tracking, measurement, and discriminative decision-making25, and it allows computers to mimic the visual ability of humans to read image content, understand dynamic scenes, and automatically extract hierarchical semantic concepts, as well as spatial and temporal associations among multi-semantic concepts embedded in visual data (e.g., images and videos)26. Convolutional neural networks (CNNs) are a class of feedforward neural networks that contain convolutional computations with a deep structure and are one of the representative algorithms of deep learning (DL)8. Since the emergence of the earliest CNNs in the 1980s and 1990s, time-delay networks27 and LeNet-5 have developed rapidly and been applied to computer vision, natural language processing, and other fields based on the advancement of computing technology28. CNNs mimic the visual perception mechanisms of living organisms and can perform both supervised and unsupervised learning. The sharing of convolutional kernel parameters within implicit layers and the sparsity of interlayer connections allow CNNs to learn grid-like topology features (e.g., pixels and audio) with less computational effort and reduced feature engineering requirements. The Massachusetts Institute of Technology (MIT) SceneParse150 benchmark provides a standard training and evaluation platform for algorithms using DL image semantic segmentation models for scene parsing, which segment and parse an image into different image regions that are related to semantic categories (e.g., sky, road, people, and bed)29. Currently, the application of AI machine image processing techniques to the quantification of color in historic urban landscapes is rare.
Despite the rapid development of AI technology, human intuitive feelings regarding landscapes will always be an important consideration in landscape evaluation, and such feelings represent an evaluation method that cannot be replaced by computer technology. There are four main methods that focus on subjective evaluation that are commonly used in academia: (1) the analytical hierarchy process, (2) Scenic Beauty Estimation (SBE), (3) psychophysiological interactions, and (4) aesthetic evaluation measurement. Of these methods, the most widely used is SBE30. The SBE method, which was formally proposed by Daniel and Boster in 1976, is based on the psychophysical school and was initially applied to the evaluation of large-scale landscapes, forested landscapes, and stand aesthetics. The SBE method has also been widely used to evaluate garden landscapes. Many scholars use the SBE method to study parks or urban landscapes and analyze the relevant factors affecting their aesthetic quality (e.g., scale appropriateness, sense of hierarchy, and color). Therefore, the SBE method can provide a theoretical foundation and basis for the study of landscape quality enhancement31,32,33,34,35.
Based on current research dynamics domestically and abroad, color-related studies are becoming increasingly detailed and systematic. However, few studies have correlated quantitative and subjective evaluations of landscape color36,37,38,39. Color is the most observable element of heritage landscape protection and planning, and how to be more rational and scientific in heritage landscape protection and planning is an important topic for heritage landscape protection. In this study, using the Yangzhou Ancient Canal as an example, five important sites of it were selected, and winter and summer landscape photographs were collected. First, by using a DLscene-parsing algorithm, quantitative color parsing was performed to extract quantitative indicators of environmental colors. Additionally, SBE was conducted to evaluate the degree of beauty of the landscape. Correlation analysis was then conducted on the quantitative color indicators and SBE values. Based on the results, we present suggestions for the color protection and planning of the canal heritage landscape.
Research aim
This study taked the Yangzhou ancient canal as the research object, extracted the quantitative color indicators by using the DL scene-parsing algorithm, analyzed the color characteristics of the canal landscape, and studied the evaluation of landscape aesthetics by scenic-beauty-estimation (SBE) method, further analyzed the correlation between the color characteristics and the SBE value, drawed conclusions and conducted analyses. This study discussed color enhancement methods suitable for canal heritage landscapes and provided color planning suggestions for the sustainable development of canal heritage landscapes.
Materials and methodology
Overview of the study area
The China’s Grand Canal is an ancient project that was added to the World Heritage List on June 22, 20147,40. The Yangzhou Ancient Canal located in Yangzhou city with a total length of approximately 30 km is an important area of the China’s Grand Canal. Based on the spatial and temporal characteristics of Yangzhou Canal development, three important crossings of the Yangzhou Canal and two areas with the greatest historical value were selected as representative landscape sites for evaluation (Fig. 1).
Location map of the study area.
(1) Site 1: The Dongguan Gudu (DGGD) was one of the most important piers of the canal in the urban section of Yangzhou during ancient times.
(2) Site 2: Gaomin Temple (GMS) is located between Guazhou and DGGD.
(3) Site 3: Guazhou Gudu (GZGD) is located at the confluence of the Chang Jiang and Grand Canal. It has been one of the most important passages from the Grand Canal into the Chang Jiang since its opening.
(4) Site 4: Canal Three Bays (YHSW) is located in the southern part of the ancient canal and is the most characteristic section of the Yangzhou ancient canal.
(5) Site 5: Yangzijin Gudu (YZJGD) is where the ancient canal enters the Chang Jiang.
Technical lines of research
In this study, five important sites of the Yangzhou Ancient Canal were selected to collect typical winter and summer photographs, and the contents and steps of the study consist of three parts: in the first part, image segmentation according to the landscape elements (sky, water, buildings, plants, and ground) was performed by using the DL scene- parsing algorithm to get the quantitative data of color and to analyze the color characteristics from the four aspects(color composition, color luminance, color saturation, and color diversity); in the second part, the evaluation of the canal landscape aesthetic of five sites in winter and summer was carried out through the SBE method by questionnaire survey; in the third part, the correlation analysis of landscape color characterization values and the SBE values was established, then the correlation conclusions and the analysis were drawn.We processed and analyzed the data according to the technical routes illustrated in Fig. 2.
Technical roadmap.
Landscape photograph data acquisition
Typical landscape scenes were selected at Site 1(DGGD), Site 2(GMS), Site 3(GZGD), Site 4(YHSW), and Site 5(YZJGD). Four angles, with at least five photos per angle, were captured using a SONY full-frame digital camera(Alpha 7R II, Sony Group, Japan) with a fixed-focus lens with a focal length of 28 mm (SEL28F20m, Sony Group, Japan). This Sony camera has 42.4 effective megapixels and a built-in five-axis stabilization system with 399 phase detection focus points. Photographs were captured under the typical summer(June 20th) and winter(December 23th) lighting conditions. Colorchecker Passport Photo2 color cards(X-Rite, USA) were used for color control correction during the photo acquisition process. Files were saved in DNG format, and color control was performed using Agfa’s ColorTune color management system. Mainly used for color calibration, matching, and color space conversion of display devices and output devices to ensure color consistency and accuracy among different devices.
Yangzhou has a subtropical monsoon climate. In spring (from March to May), the temperature rises from around 5 °C to 20–25 °C, with precipitation ranging from 200 to 300 millimeters. It’s windy and there’s a large temperature difference. In summer (from June to August), the average temperature is 20–30 °C and the precipitation is 400–600 millimeters. It’s hot and humid. In autumn (from September to November), the average temperature drops to 8–25 °C and the precipitation is 150–250 millimeters. It’s pleasant and cool. In winter (from December to February), the average temperature is 4–5 °C and the precipitation is 100–150 millimeters. It’s cold and dry (Fig. 3).
Yangzhou Average Temperature Chart(Image source: https://data.cma.cn/data/weatherBk.html).
Theory and calculation
Image segmentation
Scene parsing is the process of segmenting and parsing an image into different image regions related to semantic categories (e.g., sky, road, and people). The MIT SceneParse150 benchmark test provides a standard training and evaluation platform for scene-parsing algorithms. This benchmark test is based on data from the ADE20K dataset, which contains more than 20,000 scene-centric images containing different objects that are thoroughly annotated. The dataset contains 150 semantic categories, including areas (e.g., sky, road, and grass) and discrete objects (e.g., people, cars, and buildings). The objects in the images are unevenly distributed to simulate the natural appearance of objects in everyday scenes. For each photograph, we generated a semantic segmentation mask using different model segmentation algorithms that predict the semantic category of each pixel in the photograph. The performance of the DL algorithms was evaluated based on pixel accuracy averages and joint intersection averages for all 150 semantic categories. We used semantic image segmentation software with pyQt as a graphical user interface framework, which supports four DL models: hrnetv2, mobilenetv2dilated-c1, resnet18dilated-c1, and resnet18dilated-ppm. After testing and comparing the different models, we selected the hrnetv2 model for image processing because it provided the best segmentation effect. We segmented each photograph into five elements according to the categorization of landscape elements: sky, ground, buildings, plants, and water. After segmentation, data processing was performed to obtain the percentage of each landscape element, extract the color indicator factor of each element, and conduct statistical analysis on the color indicators.
Color extraction and quantification
Color quantification implies the simplification of color composition. The three elements of color are hue, saturation, and brightness. Hue is used to distinguish between different colors, in this study, hue information was used to define the basic color of the canal landscape as well as to determine the overall environmental color compositions. Saturation refers to the vividness of the color, also known as purity, with the higher saturation representing the more vibrant color. Brightness refers to the shade of a color, with low values representing darker shades. The HSB model can better reflect the color perception by human eyes41,42, so this study adopts the HSB model(H-Hue; S-Saturation; B-brightness). Saturation and brightness are expressed as percentages (0–100%), whereas hue is expressed as an angle (0° to 360°). Color extraction was performed using ColorImpact software. The ColorImpact software (4.2.5.707 version, Tiger Color Company, Norway ) was used for color extraction, and the following formulae were used for normalization.ColorImpact is a color scheme design tool for Windows that has won multiple awards. It combines ease of use with advanced features and provides excellent color integration for various design, multimedia, and web development programs. Normalize according to the following formula:
where h signifies hue with a value range of [0, 360], s signifies saturation with a value range of.
[0, 100], and b signifies brightness with a value range of [0, 100].
Color composition
The color composition factor was determined using the aforementioned color attributes based on the HSB model. Color composition and diversity indices were analyzed and calculated as follows.
(1) Saturation ratios refer to the ratios of extremely low-, low-, medium-, and high-saturation pixels in the color extracted from an image over the total number of pixels following normalization and are used to describe an image as a whole. A saturation ratio is calculated as
where \(\:N{S}_{i}\) is the total number of pixels in the \(\:i\)-th saturation level in the image and \(\:N\) is the total number of pixels in the image.
(2) Brightness ratios refer to the ratios of extremely low-, low-, medium-, and high-brightness pixels over the total number of pixels following normalization and are used to describe the brightness of an image. A brightness ratio is calculated as
where \(\:N{V}_{i}\) is the total number of pixels in the \(\:i\)-th brightness level in the image and \(\:N\) is the total number of pixels in the image.
Color diversity index
The color diversity index is a statistical tool used to describe the color diversity of a landscape. This index was calculated based on the number of colors extracted from the photos. The more colors in an image, the richer the image is considered to be. The color diversity index is calculated as
where \(\:{P}_{i}\) is the proportion of the \(\:i\)-th color, \(\:S\) is the number of colors.
Estimation of landscape beauty
The SBE method and questionnaire were used to eatimate the landscape beauty of the Yangzhou Ancient Canal. The evaluation steps were as follows.
-
(1)
We used the WJX.cn web platform to conduct an online questionnaire survey from February 5–12, 2023, which lasted for one week. The photos and numbers used for the questionnaire are listed below(Fig. 4):
The photographs for questionnaire survey.
-
(2)
A seven-point scale was used, with values of − 3 to 3 representing strongly dislike, dislike, somewhat dislike, average, somewhat like, like, and strongly like, respectively.
-
(3)
Evaluators browsed the questionnaires and scored the photographs in order.
-
(4)
The questionnaires were collected, checked, and collated, and SBE values were calculated.
To minimize the impact of the aesthetic preferences of individuals on the evaluation results, it was necessary to standardize the beauty degree values using the traditional standardization method.
Questionnaire production and participants
We completed the distribution and collection of questionnaires from January to February of 2023 and July to August of 2023, respectively. We designed 17 questions using the seven-point scale described above. The questionnaire was distributed electronically to 358 people with either a landscape architecture education background or non-landscape architecture education background who had received different levels of education. The participants independently assigned scores to the winter and summer landscapes of the five nodes of the Yangzhou Canal.
SBE data processing
The SBE method is the Scenic Beauty Estimation Method proposed by Daniel and Boster in 1976, and it is the most common psychophysical model in the quality evaluation methods of visual landscapes43.Data processing was performed based on the standard formulas of the SBE method presented below, and the confidence coefficient method was used to analyze the raw data in the IBM SPSS Statistics software(26 version, IBM Corporation, USA) for the overall evaluation of the winter and summer landscapes of the five nodes.
Zij: Standardized score value of the i-th judge for the i-th landscape.
Rij: Rating of the i-th judge of the j-th landscape.
Rj: Mean of the i-th judge’s ratings of all landscapes.
Sj: Standard deviation of the i-th judge’s ratings of all landscapes.
Zi: Standardized score for the i-th landscape.
Correlation analysis between quantified values of color and SBE
The SPSS software was used to analyze the bivariate correlation between the quantified color data and SBE values, and the correlation coefficients between color saturation, color diversity, and SBE values were analyzed.
Results and analyses
Color composition and characterization
The photographs were processed with image segmentation to distinguish five elements: (1) buildings, (2) sky, (3) plants, (4) water, and (5) ground. Color extraction was carried out separately for each element (as shown in Table 1). The quantification results of the winter and summer landscape colors of the five sites in the Yangzhou Ancient Canal are presented in Table 2.
As can be seen from the table, the overall Hue-values of the five sites in winter and summer are not much different. The dominant color in both winter and summer is blue and green (green: 75 ≤ h < 155; cyan blue: 155 ≤ h < 195; blue: 195 ≤ h < 275) (Table 2). The sky color at each site is blue in both winter and summer (blue: 195 ≤ h < 275), and the building color is mainly orange and yellow in both winter and summer (orange: 23 ≤ h < 50; yellow: 50 ≤ h < 75) (Table 1).
The colors of plants and water are different in winter and summer. The dominant plant color in five sites is green in summer (green: 75 ≤ h < 155), whereas in winter, changes to yellow (h: 60; yellow: 50 ≤ h < 75) in Site5(YZJGD ), or cyan (155 ≤ h < 195)in Site3:GZGD(h: 160) and Site1:DGGD (h: 180) (Table 1). This may be due to the planting of deciduous plants.
The water colors in Site1:DGGD (winter h: 100; summer h:208), Site2:GMS (winter h: 90; summer h: 209), Site4:YHSW (winter h: 120; summer h:207)and Site5:YZJGD (winter h: 140; summer h: 213)are blue (blue: 195 ≤ h < 275) in the summer, and changes to green (green: 75 ≤ h < 155) in winter.
The ground color is blue in summer (blue: 195 ≤ h < 275; summer h:207), and becomes green in winter (green: 75 ≤ h < 155). The ground, because of the small proportion of the whole photograph, has little influence on the overall color.
Color saturation analysis
We normalized the saturation values of the landscape images. Tables 3, 4, 5 and 6 present the results(Tables 3, 4, 5 and 6). The overall color saturation is found to be low in winter. The percentages of very Low Saturation in Site1(DGGD), Site2(GMS), Site3( GZGD), and Site5(YZJGD) reach 80–98%. However, Summer has Medium Saturation colors with the percentages of Mid Saturation in Site1(DGGD), Site2(GMS), Site3( GZGD), and Site4(YHSW) are 11.7%, 32.23%, 44.64% and 31.85% respectively. By category, in winter, the percentage of Very Low Saturation of building, plants, sky and water in all research sites exceeds 50%. However, in summer, the percentage of Mid Saturation of sky in Site 2(GMS) is 44.87%, which is the largest in all landscape elements. Moreover, except plant and water in Site 3(GZGD), and sky in Site2(GMS), Site 4(YHSW), the percentage of Very Low Saturation and Low Saturation of almost each landscape element is around 50% .
Brightness analysis
Analysis of the brightness of object categories (ground, building, water, sky, plants) of individual images for each studied sites (Tables 7 and 8) revealed that all the sites(DGGD, YHSW, GMS, GZGD, and YZJGD) have low brightness in winter and higher brightness in summer, sky is the only category which has exceeded 50% High Brightness in both seasons. Furthermore, in winter, the percentage of Very Low Brightness of plant in Site 2(GMS) and Site 5(YZJGD) is 24.55% and 8.26% respectively, other categories are most prevalent in Low Brightness and Medium Brightness. In summer, the percentage of Medium Brightness and High Brightness increases. For example, the percentage of High Brightness of building in Site 5(YZJGD) is 47.47%, the percentage of Mid Brightness of building in Site 1(DGGD) and Site 3(GZGD) are 91.58% and 90.65% respectively; and the percentage of Mid Brightness of plant in Site 1(DGGD) and Site2(GMS) are 92.98% and 91.14% respectively.
Color diversity analysis
The color diversity analysis of the images (Figs. 5 and 6) revealed that the color diversity of the sites in winter varies significantly, whereas the color diversity of the sites in summer varies slightly. In winter, the color diversity of Site1(DGGD) is higher than that of the other attractions. In particular, the color diversities of the water surface and plants are higher than those of the other categories (p < 0.05). The lowest color diversity is found in the buildings of Site2:GMS (p < 0.05). The color diversities of the buildings of Site1(DGGD) and Site5(YZJGD) are higher than those of the other areas (p < 0.05). The highest color diversity of plants is found in DGGD(Diversity > 1.5). That is because in summer, a lot of fall foliage plants such as Ginkgo biloba, Metasequoia glyptostroboides and Camphora officinarum are planted result in a monochromatic green color, however, in winter, with their leaves turning yellow or red, it presents a higher color diversity.In addition in summer, the difference in overall landscape diversity is insignificant (p > 0.05), with the building color diversity of Site2(GMS) being higher than that of the other sites, plant diversity of Site3(GZGD) being higher than that of the other sites (p > 0.05), and building diversity of Site5(YZJGD) being lower than that of the other sites (p < 0.05).
Color diversity index of each study site (winter).
Color diversity index of each study site (summer). Note: Due to strong light, ground reflection, plant shielding, and other factors, the graphs do not reflect the ground part.
SBE evaluation results
Questionnaire reliability analysis
The Kaiser–Meyer–Olkin (KMO) value and alpha reliability coefficient of the questionnaire results were calculated with reference to the reliability coefficient evaluation criteria of Chan et al. with the common KMO metrics: 0.9 or more means very good fit; 0.8 means good fit; 0.7 means fair fit; 0.6 means not very good fit; and 0.5 or less means very bad fit. The closer the value is to 1, the stronger the correlation between the variables41. In this study, the KMO value of the questionnaire data was 0.804, which is greater than 0.6, and the value of the reliability coefficient was 0.938, which is greater than 0.9, indicating that the data have high reliability and that the questionnaire contained high-quality information suitable for future analyses.
Basic SBE values of the five sites of the Yangzhou ancient Canal
The Scenic Beauty Estimation (SBE)values of the five sites in winter and summer were calculated using the standard formula of the Scenic Beauty Estimation (SBE) method (Table 9).
In winter, the highest SBE value is found in Site2:GMS(0.5956), and the lowest SBE value is found in Site5:YZJGD(− 0.8216), representing a difference of 1.4172, which is a large gap. The SBE values in Site2(GMS) and Site3(GZGD) which are greater than zero, indicat that these two landscapes are more desirable. The SBE values in Site5(YZJGD) which are less than zero, indicat that this landscape is the least desirable. Additionally, the SBE values also indicate that landscape in summer of Site4:YHSW(0.1127) is better than in winter(-0.3012), and the landscape of Site1:DGGD(0.1475) in winter is better than in summer(-0.1664). The highest average SBE value is in Site2:GMS(0.5062), followed by Site3:GZGD (0.2091), which both have average values greater than zero. The average SBE values of Site1:DGGD(-0.0094), Site4:YHSW(-0.0942) and Site5:YZJGD (-0.6117) are all less than zero.
Correlation analysis between quantitative color indicators and SBE values
In order to convert the saturation data in form of percentage into calculable data, we assigned the different kinds of saturation including Very Low Saturation、Low Saturation、Medium Saturation and High Saturation to numerical value − 2, -1, 1 and 2 respectively, and multiplied them by different saturation ratios to get the total saturation value(Tables 10 and 11). The bivariate correlations between color saturation and SBE values were analyzed using the SPSS software (Tables 12 and 13).
Correlation Beteween color saturation and the SBE value
The color saturation indices range from − 2.000 to − 0.403, with significant variations. Correlation analysis reveals that the p-values between the saturation and SBE values are 0.259(winter) and 0.198(summer) respectively, which are both greater than 0.05, indicating no significant correlation exists (Table 14). However, the Pearson correlation coefficients are − 0.625(winter) and 0.689(summer)respectively, indicating strong correlation. Therefore, it can be concluded that there is a strong correlation between the color saturation and SBE value in both winter and summer. It’s important to know that the color saturation in winter shows a negative correlation with SBE value, while in summer, it is a positive correlation.That means the lower color saturation in winter, the higher SBE value is, and the higher color saturation in summer, the higher SBE value is.Although the saturation index is strongly correlated with SBE value, P-value shows that this effect is not significant.
Correlation Beteween color diversity and the SBE value
Through standardized assignment, quantitative data on the color diversity of each node were obtained as shown in Tables 15 and 16. The bivariate correlation between color diversity and SBE values was analyzed using the SPSS software package (Tables 17, 18 and 19). The color diversity indices of the five nodes range from 3.178 to 7.617 in winter and summer, respectively, with large variations. Correlation analysis reveales that the p-values between the color diversity and SBE values are 0.099(winter) and 0.147(summer)respectively, which are greater than 0.05, indicating no significant correlation. However, the Pearson correlation coefficients are 0.807(winter) and − 0.747(summer)respectively, indicating strong correlation. Furthermore, we also find that the correlation trends are opposite in winter and summer. In winter, the color diversity is positively correlated with the landscape beauty, whereas in summer, the color diversity is negatively correlated with the landscape beauty.That means in winter the higher color diversity, the higher SBE value is, and in summer the lower color diversity, the higher SBE value is. Although the diversity is strongly correlated with SBE value, P-value shows that this effect is not significant.
Discussion
The deep learning scene-parsing algorithm has positive significance for landscape color research
Due to the complexity of landscape color, together with the richness of the composition of the environment, the objective research of landscape color has always been difficult. From using colorimetric color cards and color meters to using the camera to take photographs, the way of obtaining color information of landscape has become more and more accurate and objective. However, when processing photos, many scholars directly used ColorImpact, MatLab and other softwares to extract and analyze them1,2,3,4,7,8,9,17,18, which made it difficult to study the color characteristics of individual landscape elements in depth. In this study, we used DL scene-parsing algorithm to segment the image data into different landscape elements (sky, plants, buildings, water bodies, ground) through image segmentation, and extracted the color information separately. This is of positive significance for the in-depth study of landscape color.
Correlation between color characteristics and the landscape beauty
The study established a correlation between the quantitative index of landscape color and the estimation of the landscape beauty. The conclusions illustrate at least three points: ①.For the Grand Canal heritage landscape, the color saturation and diversity are strongly correlated with the landscape beauty, that means the change of color saturation and diversity directly affects the value of the landscape beauty; ②. The trend of the correlation between color saturation and landscape beauty, as well as the correlation between color diversity and landscape beauty, was influenced by seasons. This suggests that people’s preferences for landscape enjoyment are seasonally related, with people preferring less bright but richer colors in winter, and more uniform but bright colors in summer because of strong sunlight. This finding is consistent with most current research, but it is also somewhat novel. For example, some scholars have concluded that.
brighter and more colorful of plants in autumn or winter, more popular the landscape is1,18. But most of the scholars’ studies were about one season (autumn or winter).In this study, both winter and summer landscapes were studied.However, this study not only concludes that people’s preference for the landscape beauty is related to the season, but also concludes that in winter, color saturation is negatively correlated with the landscape beauty, and the diversity is positively correlated with the landscape beauty, while in summer, color saturation is positively correlated with the landscape beauty and diversity is negatively correlated with the landscape beauty.This conclusion is to some extent scientifically novel; ③.The effect of color diversity and saturation on landscape beauty is not significant. People’s evaluation of the landscape beauty is influenced by many factors, such as the combination and the pattern of the landscape, so although color saturation and diversity affect the beauty of the landscape, the effect is not significant.
Canal landscape color planning should focus on the protection and continuation of the historical color of the Canal
Originally built as a waterway for military shipping and goods transportation, the Grand Canal is a living heritage. Its original landscape features were only water and boats, with wharves and streets gradually appearing. Now, there are plant landscapes, historic districts, buildings, etc. Historically, the color style of the canal has been the main color of the water. Therefore, in heritage protection and landscape design, we should pay attention to the protection of the historical color of the canal and should not add excessively noisy color. Currently, the color composition of the Yangzhou Canal nodes is blue and green, which maintains the color tone of historical canals and protects the color memory of the canal heritage.
Planting plans for Canal landscapes should consider seasonal variations and pay attention to color saturation and color diversity
Our color quantification and correlation analysis indicates that the plant design in canals should consider seasonal changes, reduce the color saturation of plants in winter, and enhance color diversity. Simultaneously, the color saturation of summer plants should be increased while reducing color diversity. Deciduous plants occupy a large portion of the canal, resulting in a simple plant landscape with low diversity in winter. To enhance the winter landscape, colorful leafy plants could be included to enrich the color diversity and enhance the vitality of the canal. In summer, bright plants with high color saturation should be planted, but color diversity should remain low to avoid messy colors that affect the beauty of the canal heritage landscape.
Shortcomings of the study
Our research was severely hampered when the COVID-19 Pandemic was spreading.Due to the inconvenience of research, there was a limited choice of study sites and seasons.Only five sites on the Yangzhou Ancient Canal in winter and summer were considered in this study, and the limitations of the sample data leads to some limitations in the study’s conclusions.
However, this study is a meaningful attempt, and in the follow-up study, the sample range will be expanded including more addresses and other seasons, and the indicators for quantifying landscape colour will be enriched to obtain more detailed colour evaluation.Meanwhile, more samples would lead to deeper research and more valuable conclusions about the relationship between colour characteristics and landscape beauty.
Conclusions
This study chose the Yangzhou Ancient Canal as the object of study, and selected five important sites on it to study the landscape color characteristics and the relationship between colour characteristics and the landscape beauty.The DL scene parsing algorithm was applied to conduct quantitative research on landscape color. The method is used to segment landscape photographs into different landscape elements to facilitate the extraction of colour quantification indexes.Compared to previous studies on landscape colour1,17,18,19, this study allows for further detailed analysis of colour at the level of different landscape element(water, sky, building, plant and ground), thus it promotes in-depth research on landscape colour quantification.
In this study, not only landscape colour characteristics were quantified, but also the correlation between the landscape colour characteristics and the landscape beauty were further established.This represented a novel direction in the field of landscape color research. The results show that although it is statistically insignificant, there is a strong correlation between the color diversity and the SBE value, as well as the color saturation and the SBE value, and the trends of the correlation were different in winter and summer.Based on the results of data processing and analysis, some recommendations will be useful for color planning of the heritage landscapes .
In the future, the study will expand more in terms of the time and location of the samples in order to provide a more in-depth and detailed study of landscape colour and promote systematic and scientific research on the color of heritage landscapes.
Data availability
The datasets generated and/or analysed during the current study are not publicly available due to being self-researched and privately owned, but are available from the corresponding author on reasonable request.
References
Cheng, Y. & Tan, M. The quantitative research of landscape color: A study of Ming dynasty City wall in Nanjing. Color. Res. Appl. 43, 436–448. https://doi.org/10.1002/col.22203 (2018).
Shuttleworth, S. The use of photographs as an environmental presentation medium in landscape studies. J. Environ. Manag. 11, 285–305 (1980).
Daniel, T. C. Whither scenic beauty? Visual landscape quality assessment in the 21st century. Landsc. Urban Plan. 54, 267–281. https://doi.org/10.1016/S0169-2046(01)00141-4 (2001).
Villafuerte, N. Digital imaging for colour measurement in ecological research. Ecol. Lett. 1, 151–154. https://doi.org/10.1046/j.1461-0248.1998.00034.x (1998).
Kendal, D. et al. Quantifying plant colour and colour difference as perceived by humans using digital images. PLOS ONE. 8, e72296. https://doi.org/10.1371/journal.pone.0072296 (2013).
Huang, H. et al. Clusters induced electron redistribution to tune oxygen reduction activity of transition metal single-atom for metal-air batteries. Angew Chem. Int. Ed. Engl. 61, e202116068. https://doi.org/10.1002/anie.202116068 (2022).
Li, X. China’s grand Canal geographical names culture enters the united nations the beautiful scenery of the grand Canal: Yangzhou Beijing-Hangzhou Canal Bridge. Civil Affairs China. 18, 66. https://doi.org/10.25236/FSST.2020.020107 (2021).
Bao, C. L. & Chen, Y. Research on application of convolutional neural network in remote sensing image recognition, (2020). https://doi.org/10.23977/mcee2020.017
Hou, B. J. & Hu, Y. Research on the local response of the cultural construction of the grand Canal from the perspective of integrated development of culture and tourism: A case study of cities along the Jiangsu-Zhejiang section of the grand Canal in China. China Famous Cities. 35, 50–59. https://doi.org/10.19924/j.cnki.1674-4144.2021.09.008 (2021).
Zhang, L. et al. A quantitative study on colour of city landmark landscape architectures, A Quantitative Study on the Colour of City Landmark Landscape Architectures, J. Phys. Conf. 1288 (2019). https://doi.org/10.1088/1742-6596/1288/1/012011
Xing, X., Hao, P. & Dong, L. Color characteristics of Beijing’s regional Woody vegetation based on natural color system. Color. Res. Appl. 44, 595–612. https://doi.org/10.1002/col.22375 (2019).
Huang, Q. & Chen, F. H. Comparative analysis of four major color systems. Packag Eng. 40, 266–272. https://doi.org/10.19554/j.cnki.1001-3563.2019.06.043 (2019).
Sun, Y. M. Color Quantization and Evaluation Research on Fall-Color Trees in Beijing (Beijing Forestry University, 2015).
Kasajima, I. Measuring plant colors. Plant. Biotechnol. (Tokyo). 36, 63–75. https://doi.org/10.5511/plantbiotechnology.19.0322a (2019).
Xie, Y., Lin, T., Chen, R. & Chen, Z. Toward improved aesthetics and data discrimination for treemaps via color schemes. Color. Res. Appl. 43, 328–340. https://doi.org/10.1002/col.22196 (2018).
Cohen, D. A. et al. The potential for pocket parks to increase physical activity, Am. J. Health Promot. 28 (3) Supplement S19–S26. (2014). https://doi.org/10.4278/ajhp.130430-QUAN-213
Shen, S., Yao, Y. & Li, C. Quantitative study on landscape colors of plant communities in urban parks based on natural color system and M-S theory in Nanjing, China. Color. Res. Application. 47, 152–163. https://doi.org/10.1002/col.22713 (2022).
Zhang, X. J., Chen, J., Li, Q. Y., Liu, J. C. & Tao, J. P. Color quantification and evaluation of landscape aesthetic quality for autumn landscape forest based on visual characteristics in subalpine region of Western Sichuan, China. Ying Yong Sheng Tai Xue Bao. 31, 45–54. https://doi.org/10.13287/j.1001-9332.202001.016 (2020).
Yang, J., Wang, X. R. & Zhao, Y. Leaf color attributes of urban colored-leaf plants. Open. Geosci. 14, 1591–1605. https://doi.org/10.1515/geo-2022-0433 (2022).
Saeedi, I. & Dabbagh, E. Modeling the relationships between hardscape color and user satisfaction in urban parks. Environ. Dev. Sustain. 23, 6535–6552. https://doi.org/10.1007/s10668-020-00866-z (2021).
Cesar, J. C. O. Chromatic harmony in architecture and the Munsell color system. Color. Res. Appl. Wiley Hoboken. 43, 865–871. https://doi.org/10.1002/col.22283 (2018).
Li, K. R., Yang, Y. Q. & Zheng, Z. Q. Research on color harmony of Building façades. Color. Res. Appl. 45, 105–119. https://doi.org/10.1002/col.22448 (2020).
Zhang, L., Cheng, Y. & Ma, L. A quantitative study on the colour of City landmark landscape architectures. J. Phys. : Conf. Ser. 1288, 012011. https://doi.org/10.1088/1742-6596/1288/1/012011 (2019).
Qin, X., Zhang, N., Zhang, W. & Meitner, M. How does tunnel interior color environment influence driving behavior? Quantitative analysis and assessment experiment, tunnelling undergr. Space Technol. 98, 103320. https://doi.org/10.1016/j.tust.2020.103320 (2020).
Liu, J., Ji, K. & Jing, Y. Image Recognition and Extraction of Students, Human Motion Features Based on Graph Neural Network. Comput. Intell. Neurosci. 2022, 6755053. https://doi.org/10.1155/2022/6755053 (2022).
Bouguettaya, A., Zarzour, H., Kechida, A. & Taberkit, A. M. Deep learning techniques to classify agricultural crops through UAV imagery: A review. Neural Comput. Appl. 34, 9511–9536. https://doi.org/10.1007/s00521-022-07104-9 (2022).
Wang, S., Yu, M. & Sun, X. Robust h ∞ control for time-delay networked control systems with probability constraints. IET Control Theor. Appl. 9, 2482–2489. https://doi.org/10.1049/iet-cta.2015.0143 (2015).
Chew, R. et al. Deep neural networks and transfer learning for food crop identification in UAV images. Drones 4, 7. https://doi.org/10.3390/drones4010007 (2020).
Sun, K., Xiao, B., Liu, D. & Wang, J. Deep High-Resolution representation learning for human pose Estimation. IEEE Publications. 5686–5696. https://doi.org/10.1109/CVPR.2019.00584 (2019).
Akbar, K. F., Hale, W. H. G. & Headley, A. D. Assessment of scenic beauty of the roadside vegetation in Northern England. Landsc. Urban Plan. 63, 139–144. https://doi.org/10.1016/S0169-2046(02)00185-8 (2003).
Li, H., Shi, K., Wang, Y., Li, Y. & Feng, Y. Research on scenic beauty Estimation of plant landscape on the roof on SBE method. Arab. J. Geosci. 14, 882. https://doi.org/10.1007/s12517-021-07225-w (2021).
Shi, Y., Lv, D. & He, J. Landscape evaluation of urban parks based on SBE and AHP: A case study of Kunming City. IOP Conf. Ser. : Earth Environ. Sci. 580, 012016. https://doi.org/10.1088/1755-1315/580/1/012016 (2020).
Mei, G. & Sun, Y. Using landscape preference models and tree-scale measurements to predict stand quality. Scand. J. For. Res. 33, 255–260. https://doi.org/10.1080/02827581.2017.1367839 (2018).
Huang, Y., Li, T., Jin, Y. & Wu, W. Correlations among AHP-based scenic beauty Estimation and water quality indicators of typical urban constructed WQT wetland park landscaping. AQUA — Water Infrastructure Ecosyst. Soc. 72, 2017–2034. https://doi.org/10.2166/aqua.2023.071 (2023).
Liu, S., Sun, Y., Ouyang, H. & Ma, X. A study on the landscape evaluation of sculpture vignettes in residential areas based on SBE. Appl. Math. Nonlinear Sci. 9 https://doi.org/10.2478/amns.2023.2.00128 (2024).
Pérez-Delgado, M. L. & Celebi, M. E. A comparative study of color quantization methods using various image quality assessment indices. Multimedia Syst. 30 https://doi.org/10.1007/s00530-023-01206-7 (2024).
Feng, J. et al. Quantitative study on color characteristics of urban park landscapes based on K-means clustering and SD. Earth Sci. Inf. 17 999–1012. https://doi.org/10.1007/s12145-024-01235-4. (2024).
Shao, J., Li, L., Jiang, Y., Liu, X. & Yang, H. Quantitative analysis methods for evaluating colour on architectural heritage: survey of colour restoration and perception of the Wen Yuan ge in the forbidden City. Herit. Sci. 12, 124. https://doi.org/10.1186/s40494-024-01218-0 (2014).
Zhang, C. & Tan, G. Quantifying architectural color quality: A machine learning integrated framework driven by quantitative color metrics. Ecol. Indic. 157, 111237. https://doi.org/10.1016/j.ecolind.2023.111237 (2023).
Zhang, Y. et al. Habitat quality assessment and ecological risks prediction: an analysis in the Beijing-Hangzhou grand Canal (Suzhou section), water. 14 2602. (2022). https://doi.org/10.3390/w14172602.auto-
Chen, T. W., Chen, Y. L. & Chien, S. Y. Fast image segmentation based on K-Means clustering with histograms in HSV color space [C]//2008 IEEE 10th Workshop on Multimedia Signal Processing. IEEE, : 322–325. (2008).
Smith, J. R. & Chang, S. F. Integrated Spatial and feature image query [J]. Multimedia Syst. 7 (2), 129–140 (1999).
Daniel, T. C. & Boster, R. S. Measuring Landscape Esthetics: the Scenicbeauty Estimation on Method [D] (USDA Forest Service, 1976).
Author information
Authors and Affiliations
Contributions
Yanyan Wang and Jianglin Qian wrote the main manuscript text. Jiajie Cao, Rong Fan and Xunyu Han prepared the figures and tables. All authors reviewed the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethical approval and consent to participate
Ethics clearance for the study was obtained from Ethics Review Committee (ERC) of Nanjing Forestry University. All methods were carried out in accordance with relevant guidelines and regulations of the ERC. Informed consent was obtained from all subjects and/or their legal guardian(s).
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Wang, Y., Qian, J., Cao, J. et al. Quantitative analysis and evaluation of winter and summer landscape colors in the Yangzhou ancient Canal utilizing deep learning. Sci Rep 15, 7500 (2025). https://doi.org/10.1038/s41598-025-91483-1
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1038/s41598-025-91483-1
