| As a typical blue-green space in cities,lake parks provide important ecological regulation services and landscape recreation services,with various environmental and social benefits,which are beneficial to improving urban recreation environment,human well-being and public health.However,the ecological environment of lakes has been ruined in the context of urban water crisis and rapid urbanization,which affects the healthy and sustainable development of urban habitat.Therefore,protecting and improving urban lake resources and promoting the landscape quality of lake parks have become urgent issues to be solved in urban construction.From the perspective of public perception to explore the qualities of the landscape,reveal the influence mechanism of the landscape visual elements on the potential emotional tendency of visitors,which helps to create a truly "people-oriented" landscape and satisfy people’s demand for high-quality human living environment.This study takes 20 completed lake parks in the central city of Wuhan as the research object,and uses various computer vision algorithms to parametrically analyze 35,631 landscape images from social media to extract and quantify the landscape features in the images,including three major parts of image classification and recognition,image semantic segmentation and image color quantification:(i)Using Google Cloud Vision algorithm,through the six modules including data preparation,image label analysis,manual label addition,model training,model validity assessment and batch prediction,the landscape elements classification,spatial scale classification and landscape elements recognition Auto ML models were build to analyze the landscape types,spatial scales and landscape elements of the lake park.(ii)The image content was segmented by DeepLab v3+ algorithm to obtain three indicators: green view index,sky view index and building view index.(iii)Using OpenCV and K-means clustering algorithm to extract the main colors of the images and their HSV feature values to realize the colour clustering and quantification of the images.Based on the landscape features extracted by image parametric processing,three major dimensions of landscape image are summarized: landscape composition,landscape proportion and landscape color,and a multi-dimensional framework for quantitative measurement of landscape image based on public perception is developed.Furthermore,the perceptual commonality and characteristics of landscape image,the public’s preferences and the reasons affecting the differences,and the temporal and spatial change patterns in lake parks are quantitatively analyzed through various mathematical and statistical methods and spatial analysis.The main conclusions obtained from this study are as follows.(1)Landscape composition dimension: in terms of landscape types,the perception of natural landscape in lake parks is significantly higher than that of man-made landscape,among which water landscape,land plants landscape and historical and culture are the core landscape types of lake parks;In terms of spatial scale,there is no significant difference in public preference for landscapes of different scales(macro-medium-micro),and natural landscape covering a large area with a broad view and man-made landscape with delicate designs covering a relatively small area are preferred by tourists;In terms of landscape elements,evergreen trees,lake,grassland,background buildings and reflection are the representative landscape elements of lake parks;which reflect people’s aspiration for natural scenery and identification with regional history and culture.Secondly,the study shows that the differences in the public’s preference for different landscapes are significant,with idiosyncratic landscapes representing park characteristics more likely to be highly perceived by people,indicating the importance of landscape trait mining and differentiated landscape construction.In addition,people’s preference for each landscape type and landscape element in different seasons also differs to a certain extent.(2)Landscape proportion dimension: the average green view index(GVI),sky visibility(SVI)and building visibility(BVI)of the lake park are 0.3597,0.1863 and0.0741.Generally speaking,landscapes with medium green view index(0.3<GVI<0.5),medium spatial openness(0.10<SVI<0.25)and low building visibility(0.01<BVI<0.1)are more preferred by the public,except that people have different preference characteristics for the proportion of landscapes with different landscape types and spatial scales.(3)Landscape color dimension: the color of man-made landscape is more diversified compared with natural landscape,with greater differences within the group.Overall,the dominant color tone of the lake park landscape is blue-green,and presents the low saturation and medium brightness.From the perspective of color psychology,people prefer landscapes with soft colors and low visual stimulation in lakes and parks,which can make people calm and soothe.In general,the preferred landscape image of the lake park presents the characteristics of combining naturalness and historicity,coexistence of homogeneity and heterogeneity,reasonable ratio of visual elements,significant temporal and spatial change patterns,and diverse color visual effects.Based on the above findings,this study proposes suggestions for the improvement of landscape visual quality of lake parks in terms of maintaining natural authenticity,attaching importance to historical heritage,increasing landscape openness and change,mining landscape characteristics,optimizing landscape visual elements and reasonably designing landscape colors.This study enriches the analysis method and content framework of landscape preference,provides new methodological ideas and technical support for the research related to perceptual preference analysis with big data images,and has theoretical and practical application value for guiding the construction of lake parks and urban landscapes. |
