摘要
立足视觉感知分析人行横道处的道路环境特征,对提升行人安全具有重要意义。基于街景图像,从语义分割、色彩计算和深度估计3个维度自动提取视觉道路环境特征;采用层次聚类将人行横道处的视觉道路环境划分为自然主导低饱和度型、建筑主导高视觉复杂度型及开阔且均衡型3种类型。结果表明:这3种类型对应的行人事故平均数分别为0.741、0.457与0.380,其中开阔且均衡型具有最高行人安全性;与单一视觉特征相比,视觉道路环境类别与行人事故的相关性更强,更能反映驾驶人与行人对场景的整体视觉感知。
Analyzing road environment features at pedestrian crosswalks from the perspective of visual perception is of great significance for improving pedestrian safety.Based on street-view images,this study automatically extracts visual road environment features in terms of semantic composition,color attributes,and depth information Hierarchical clustering is applied to classify the visual road environments at pedestrian crosswalks into three categories:nature-dominated with low saturation,building-dominated with high visual complexity,and open-and-balanced.The result shows that the corresponding average numbers of pedestrian crashes for these categories are 0.741,0.457,and 0.380,respectively,with the open-and-balanced type exhibiting the highest level of pedestrian safety.Compared with single visual features,the categorized visual road environment types show a stronger correlation with pedestrian crashes and better capture the holistic visual perception experienced by drivers and pedestrians.
作者
任蔚溪
陈雨人
REN Weixi;CHEN Yuren(The Key Laboratory of Road and Traffic Engineering,Ministry of Education,Tongji University,Shanghai 201804,China)
出处
《交通与运输》
2026年第1期88-93,共6页
Traffic & Transportation
基金
国家自然科学基金项目(52102416)。
关键词
人行横道
视觉道路环境
视觉特征
层次聚类
街景图像
crosswalk
road environment
hierarchical clustering
visual features
street-view images
