| 摘要: |
| 机器学习使实现数据的智能化处理及充分利用数据中蕴含的知识与价值成为可能。探索基于机器学习在风景园林领域智能化分析应用的途径，开展3个实验。其中2个与数据分析研究相关，提出基于调研图像色彩聚类分析的城市色彩印象和基于图像识别技术的景观视觉质量评估与网络应用平台部署实验。最后1个实验与数字化设计创作相关，提出用于设计方案遴选的地形生成方法，包括2个子项目：应用深度学习生成对抗网络（GAN）的地形生成和建立遮罩、预测未知区域的高程。3个实验应用到机器学习中分类、聚类和回归3个主要方向中的算法以及深度学习的生成对抗网络，对传统的研究问题提出了基于机器学习新的研究方法。因此，在应用机器学习风景园林领域，可以有效地从多源数据中学习相互增强的知识，发现问题，并提出解决问题的新方法。 |
| 关键词: 风景园林 数字景观 机器学习 城市色彩 视觉评价 地形生成 |
| DOI： |
| 分类号:TU986 |
| 基金项目: |
|
| Research on Intellectual Analysis and Application of Landscape Architecture Based on Machine Learning |
|
BAO Ruiqing
|
|
Xi’an University of Architecture and Technology
|
| Abstract: |
| Machine learning makes data intellectualization processing and full use of knowledge and value contained in data possible. To explore the approaches of intellectual analysis and application in landscape architecture based on machine learning, we carried out three experiments. Two experiments are related to data analysis study, to propose urban colour impression based on investigating image colour clustering analysis, and the landscape visual quality evaluation and network application platform deployment based on image recognition technology. The last experiment is related to digital design creation, to propose the terrain generation method for design scheme comparison, including two subitems, which are the terrain generation with the Generative Adversarial Networks of deep learning as well as predicting the elevation in the unknown area by building masks, respectively. All the three experiments employ the algorithms of classification, clustering, and regression, as well as the Generative Adversarial Networks of deep learning, to propose new study methods based on machine learning for traditional study issues. Therefore, with the application of machine learning to landscape architecture, we can effectively learn about the interaction enhancement knowledge from multi-source, find problems and propose new methods to solve problems. |
| Key words: landscape architecture digital landscape machine learning urban color visual evaluation terrain generation |
