Research of Evaluation Models and Thresholds for Park Landscaping Techniques Based on Eye Movement Analysis:A Case Study Based on Borrowing Scenery and Framing Scenery Methods

XIAO Xi; ZHENG Yunfei; CAI Jinhui; HUANG Hanjie; LI Haolong; MA Yuan

doi:10.3969/j.issn.1000-0232.2025.06.010

Abstract：Landscaping techniques such as borrowing scene methods(jie jing)and framing scene methods(kuang jing)hold significant reference and guiding value for the landscape design of modern urban parks.To address the over-reliance on empirical knowledge and lack of quantitative foundations in traditional landscaping techniques,this study employs eye-tracking technology to quantitatively investigate the application efficacy,optimal scale distance thresholds,and influencing factors of borrowing scene and framing scene methods in modern urban parks.It aims to establish a quantitative evaluation system for landscape experiences,providing scientific and parametric foundations for landscape design. In this study,five representative urban parks,including Guangzhou's Liuhua Lake Park,were chosen,and 12 typical scenes(six in borrowing scenes and six in framing scenes)were identified for standardized photographic documentation,collecting 48 experimental images.A total of 14 eye-tracking data metrics of 50 participants(25 males and 25 females)were recorded using the HTC Vive Pro2 eye-tracking equipment.Meanwhile,questionnaire data on landscape satisfaction scores was collected.Coupled analyses,including Spearman correlation and independent sample t-tests,were conducted between eye-tracking data and physical spatial parameters to explore the influences of differences in participants and imagery element characteristics on landscape experiences of borrowing scenes and framing scenes.Eye-tracking indicators that influence scenic beauty were further explored by combining subjective evaluation data on questionnaires,and a mathematical model of experience evaluation was built using stepwise regression.Finally,polynomial training simulations were established by machine learning,and the predictive model of evaluation scores was fit.The distance thresholds for borrowing and framing scenes under the optimal landscaping scores were calculated. Research results showed:(1)According to gender difference analysis,there are significant differences between males and females in scenery attention and cognition.For example,females show more detailed scene cognition,whereas males show higher information processing efficiency.(2)The spatial imagery elements show that different imagery elements have different characteristics.For instance,the visual focus of point elements in the borrowing scenery method is significantly better than those of planar and linear elements.In the framing scenery method,the aspect ratios and frame proportions can affect visual focus significantly.(3)The regression model shows that the borrowing scenery evaluation(EVQ=8.076+0.890 ×AFD-0.464 ×APD+0.244 ×TST)and the framing scenery evaluation(EVQ=7.464+0.025 × NF-0.496 × APD+0.013 ×NES)both showed significant correlations with subjective ratings(p＜0.01).(4)The optimal landscape distance thresholds were determined at 50-250 m for borrowing scenes and 1-7 m for framing scenes through machine learning simulation. Some major conclusions could be drawn.First,there might be a scientific mechanism in traditional landscaping techniques.The optimal landscape thresholds are explicit in both borrowing and framing scenes;the layout of landscape points can be optimized during design according to distance.Second,point elements and frame aspect ratios are critical for enhancing visual focus.It is suggested that landscape effects be strengthened by integrating landmark buildings and designing horizontal frameworks.Third,a subjective experience can be predicted effectively from the quantitative model of eye-tracking data,which provides scientific tools for landscape design. The perceptual thresholds and key design parameters of borrowing scenery and framing scenery methods are quantified systematically by integrating eye-tracking technology and machine learning.Recommended quantitative design parameters are provided for landscape construction in modern urban parks.Research results provide theoretical and practical references to improve landscape experience in urban parks.Limited by static experimental media and sample size,the universality of thresholds shall be verified in dynamic environments,and multiple types of scenes shall be involved in future studies.