基于图像语义分割法的城市冠顶式森林步道景观质量评价——以福州福道为例

Landscape Quality Evaluation of Urban Treetop Walk Based on Image Semantic Segmentation—Taking Fu Forest Trail in Fuzhou as an Example

城市冠顶式森林步道的构建对增进公众福祉具有重要意义,提升城市冠顶式森林步道景观质量是其建设的重要内容,而定量评价景观要素的美景度是提升景观质量的基础。以福州福道为例,首先利用灰色统计分析法预选其景观要素,然后采用图像语义分割法和景观美景度评价法对福道的景观要素进行定量分析与评价,最后通过偏相关分析,建立福道景观美景度评价的多元线性回归模型,用于福道景观质量评价。结果表明:①PSPNet对植被和天空景观要素的识别精度分别达到88.61%和90.91%,说明图像语义分割法可实现对福道景观要素的有效量化;②景观美景度值分析结果显示,山体侧有植被步道区域的总体景观美景度评分最高,休憩亭廊区域的总体景观美景度评分最低;③偏相关分析结果显示,天空开阔度、绿视率、植物色彩丰富度、最远视线距离、可视区域面积和场景色彩丰富指数是决定福道景观美景度的主要因素,其影响程度依次减小,建立的福道景观美景度评价多元线性回归模型具有良好的解释度和统计学意义。该研究结果可为城市冠顶式森林步道景观质量的定量评价提供理论依据。

The construction of urban treetop walk is of great significance for improving public welfare.Improving the quality of its landscape is an important part of the construction,and quantitative evaluation for the scenic beauty of landscape elements is the basis for improving the quality of the landscape.Therefore,this paper takes Fu Forest Trail in Fuzhou as an example,pre-selects influencing factors through the grey statistical analysis method,and explores the use of image semantic segmentation and scenic beauty evaluation to quantitatively analyze and evaluate the landscape elements of treetop walk.Finally,through partial correlation analysis,the paper establishes a multiple regression model for landscape beauty evaluation to evaluate the landscape quality of Fu Forest trail.The results are as follows:①The recognition accuracy of plant and sky landscape elements reaches 88.61%and 90.91%,respectively,indicating effective quantification of landscape elements by the image semantic segmentation method;②Fu forest trail is divided into 4 types of main line plank road space.The scenic beauty value analysis results show that the vegetation sample plot on the side of the mountain has the highest overall scenic beauty rating,and the rest pavilion the lowest;③Partial correlation analysis results show that sky openness index,green view index,plant color enrichment index,farthest line of sight distance,visible area and scene color enrichment index are the key factors that determine the beauty of Fu Forest Trail landscape,and the degree of influence decreases in order.The multiple regression model of landscape beauty has good explanatory degree and statistical significance.The research results provide a theoretical basis for the quantitative evaluation of urban treetop walk.