西北林学院学报2024，Vol.39Issue(3)：231-238,8.DOI:10.3969/j.issn.1001-7461.2024.03.29
基于PSPNet深度学习网络景观要素语义分割的春季森林景观质量评价
Evaluation of Forest Landscape Quality in Spring Based on Semantic Segmentation of Landscape Elements in PSPNet Deep Learning Network
摘要
Abstract
Based on BIB-LCJ(balanced incomplete block-law of comparative judgment)method,25 samples of forest landscape in spring were evaluated from the view of scenic beauty.We combined PSPNet(pyra-mid scene parsing network)deep learning network to conduct semantic segmentation of spring forest land-scape images and to establish a multiple linear regression model of beauty rating and landscape elements.The results showed that 1)the recognition accuracies of PSPNet for sky and vegetation were 87.9％and 86.7％,respectively,which were within the usable range.2)Among the 7 spring forest landscape elements related to the beauty degree,"plant level""flower color richness"and"color uniformity"were the most important ones."Color uniformity"could improve the beauty of spring landscape,while"visibility of dead branches""proportion of irrigation and grass layers""soil exposure"and"proportion of visual disturb-ances"had negative effects on the beauty of the landscape.The results of the study can guide the creation and quality improvement of spring forest landscape.In addition,the semantic segmentation technology pro-vided by the deep learning PSPNet model effectively solves the technical bottleneck of the application of the objective landscape element decomposition based on computer vision method of quantifying the composition of landscape images.关键词
语义分割/美景度/景观要素分解/森林景观质量评价/深度学习Key words
semantic segmentation/beauty degree/landscape element decomposition/forest landscape quality evaluation/deep learning分类
农业科技引用本文复制引用
刘帅健,邓华锋..基于PSPNet深度学习网络景观要素语义分割的春季森林景观质量评价[J].西北林学院学报,2024,39(3):231-238,8.基金项目
国家自然科学基金(72173011). （72173011）
