摘要
针对现有室内POI显著度评价多采用依赖专家评分确定权重的线性模型而难以适应环境变化的问题,提出了一种数据驱动的室内POI显著度评价方法。基于室内POI样本数据,利用随机森林模型来探究视觉、语义和结构等指标与显著度的非线性关系。实验招募了138名志愿者在某典型室内商场中开展地标认知试验,将试验中POI的回忆人数作为其显著度值;选择了包含视觉、语义和结构特征的10个指标作为随机森林模型构建的输入变量,较好地克服了线性模型的局限,实现了对特征指标的精确提取。实验结果表明:与先前研究模型相比,该方法在预测精度和模型解释能力方面均表现更好,证实了该方法能够拟合室内POI显著度与其评价指标之间的非线性关系。此外,随机森林模型对特征变量的重要性评估为研究者在指标体系构建和权重设置上提供了一定的参考。
Aiming at the problem that existing indoor POI saliency evaluations mostly use linear models relying on expert ratings to determine weights and are difficult to adapt to environmental changes,this paper proposes a data-driven method for indoor POI saliency evaluation.Based on indoor POI sample data,a random forest model is utilized to explore the nonlinear relationship between visual,semantic and structural metrics and saliency.The experiment recruited 138 volunteers to conduct a landmark awareness test in a typical indoor shopping mall,and the number of POI recalls in the test was taken as its saliency value;10 indicators containing visual,semantic and structural features were selected as input variables for the construction of the random forest model.The limitations of the linear model were better overcome,and the precise extraction of the feature indicators was realized.The experimental results show that the method performs better in terms of both prediction accuracy and model interpretation ability compared with previous research models,confirming that the method can fit the nonlinear relationship between indoor POI salience and its evaluation metrics.In addition,the random forest model's assessment of the significance of the characteristic variables provides a certain reference for researchers in the construction of the indicator system and the setting of weights.
作者
李华蓉
赵芹
李天童
梁渝清
陈团
王鑫
LI Huarong;ZHAO Qin;LI Tiantong;LIANG Yuqing;CHEN Tuan;WANG Xin(College of Smart City,Chongqing Jiaotong University,Chongqing 40o074,China;Chongqing Institute of Surveying and Monitoring for Planning and Natural Resources,Chongqing 401120,China)
出处
《测绘科学》
北大核心
2025年第5期196-206,共11页
Science of Surveying and Mapping
基金
重庆市研究生联合培养基地项目(JDLHPYJD2020005)
重庆市自然科学基金项目(CSTB2023NSCQ-MSX0880,CSTB2022NSCQ-MSX1625)
重庆市教育委员会科学技术研究项目(KJQN202100734)。
