Research On Extracting Visual Perception Indicators Of Highway Space Based On Driver’s Dynamic Field Of Vision

As an important part of the national comprehensive transportation network,the highway network has great benefits for China’s economic development and resource allocation.However,with the high-speed and large-scale growth of automobile ownership and highway mileage,traffic safety issues are becoming increasingly severe.Driver factors play a leading role in traffic accidents,and in-depth exploration of drivers’ subjective visual perceptions in highway space is crucial.Quantifying and extracting visual perception indicators from a driver’s perspective has become an urgent problem that needs to be solved.Based on the analysis of the characteristics of highway space and the driver’s dynamic visual perception theory,this paper constructed a model of the driver’s dynamic visual field range,and thus determined the dynamic visual field range of drivers under different design speeds.Based on the characteristics of visual perception in highway space and the principles of index selection,the visual perception indicators of green quantity,visual field openness,and visual information complexity in highway space were clarified,and the data of the highway network and Baidu street view images in the northeast of Beijing were obtained and processed.Using the fully convolutional neural network(FCN)semantic segmentation framework and digital image processing code as tools,the green view index,sky view factor,and visual entropy under different speeds and panorama views were quantified and compared.Based on the semantic analysis method,a subjective evaluation questionnaire was designed to deeply analyze the differences in subjective evaluation results.Finally,the effectiveness of the dynamic visual range model was verified by combining the correlation between the subjective and objective data of each individual index,and a multivariate linear regression analysis model was constructed between the indexes,completing the clustering analysis and evaluation of the overall visual perception of the highway network in the northeast region of Beijing.The results showed that there is a strong linear correlation between the green view index,sky view factor,and visual entropy under different speeds and panoramas and their corresponding green quantity,visual field openness,and visual information complexity.Moreover,the Pearson correlation coefficient between each objective and subjective data extracted based on the driver’s dynamic visual field range model is higher than that of the panorama,validating that the objective indicators extracted by considering the visual field range are closer to the driver’s subjective visual perception in the highway space.An OVP regression model of the overall visual perception in highway space was constructed,and it was found that sky view factor and green view index have a greater impact on the overall visual perception,while visual entropy has a small and negative correlation with it.The visual perception indicators of highway space extracted based on the driver’s dynamic visual range model truly realize the process of quantifying spatial visual perception from the driver’s subjective perspective.The regression model for the overall visual perception of highway space is also helpful in providing indicators that are more closely related to the real feelings of road users for future highway space planning,design,and renovation,and providing relevant guidance for highway landscape design,with high practical value for engineering.