随着科学技术的发展，在人工神经网络获得突破性进展之后，人工智能机器学习变 成一个热门的研究方向，不少传统行业都希望借助人工智能技术来替代传统人力的工作， 实现对行业天花板的突破。风景园林设计在经过数十年的数字化发展，不仅积累了大量 的数字化成果，在生产技术层面也不断革新着方法和策略，但其传统的植物配置设计环 节在很大程度上仍受限于景观设计师个人的专业知识，从业经验以及对相关生产软件的 熟练程度，而且对历史数据的复用率极低，因此在设计的科学合理性以及工作效率方面 都存在较大的提升空间。人工神经网络技术具有学习和处理海量数据的能力，适合于解 决复杂的，非线性的问题，已应用在金融、运输、医疗、军事等众多行业，理论上也可 以应用于风景园林行业。 本报告通过对人工神经网络结构、特性和运作原理的研究，结合风景园林植物配置 设计的过程和方法以及其在开放场地环境下受多种环境因子影响的复杂特性，假设了一 套将顶层设计参数、环境影响因子、植物配置方法、绿化景观效果评价体系等维度相结 合的应用模型。通过查阅相关文献，研究相似案例以及对模型部分关键步骤进行实操还 原来对人工神经网络应用在植物配植设计的可行性上进行判断，总结出两者具有高度耦 合性，具备较高的应用可能。 最后，本报告还结合人工智能的技术发展趋势和目前的前沿技术案例来探讨未来设 计生成系统的优化设想，并给出景观师在未来人工智能时代的应对策略。 关键词，人工神经网络；机器学习；自生成系统；植物配置；II Abstract With the development of science and technology, after artificial neural network breakthrough, artificial intelligence machine learning has become a popular research direction, many traditional industries want to use artificial intelligence technology to replace the traditional human work, to achieve the industry ceiling. The break. Landscape design After decades of digital development, not only accumulated a large number of digital achievements in the production technology level has also been innovative methods and strategies, but the traditional plant configuration design links to a large extent still limited to the landscape The designer's personal knowledge, experience and proficiency in the production of relevant software, and the reuse rate of historical data is very low, so the design of scientific rationality and efficiency there is a large room for improvement. Artificial neural network technology has the ability to learn and deal with massive data, suitable for solving complex, non-linear problems, has been used in finance, transportation, medical, military and many other industries, theoretically can also be applied to the landscape industry. In this paper, through the study of the structure, characteristics and operating principle of artificial neural network, combined with the process and method of landscape plant configuration design and its complex characteristics influenced by various environmental factors in open environment, it is assumed that a set of top- Parameters, environmental impact factors, plant allocation methods, green landscape effect evaluation system and other dimensions of the combination of the application model. By examining the relevant literature, studying similar cases and practicing the key steps of the model, the feasibility of artificial neural network application in plant planting design is judged. It is concluded that the two are highly coupled and highly viable. Finally, this paper also discusses the optimization design of future design generation system based on the development trend of artificial intelligence and the current cutting-edge technology case, and gives the designer's coping strategies in the future artificial intelligence era. Keywords，Artificial Neural Network; Machine Learning; Self-generating System; Plant Arrangement;III 目录 摘要 ............................................................ I ABSTRACT ........................................................... II 目 录 .......................................................... III 第一章 绪 论 ....................................................... 1 1.1 相关概念 ........................................................ 1 1.1.1 人工神经网络 ................................................ 1 1.1.2 机器学习 .................................................... 1 1.1.3 自生成系统 .................................................. 1 1.1.4 植物配置 .................................................... 2 1.2 研究背景 ........................................................ 2 1.3 研究意义 ........................................................ 3 1.4 国内外研究现状 .................................................. 4 1.5 研究目标 ........................................................ 6 1.6 研究内容 ........................................................ 6 1.7 研究方法和技术路线 .............................................. 7 1.8 研究工作基础 .................................................... 9 第二章 理论可行性研究 .............................................. 10 2.1 风景园林设计的数字化基础 ....................................... 10 2.1.1 二维辅助绘图阶段 ........................................... 11 2.1.2 三维辅助建模阶段 ........................................... 12 2.1.3 数字化生态模拟阶段 ......................................... 14 2.1.4 参数化设计与算法设计阶段 ................................... 18 2.2 人工神经网络的原理和特性 ....................................... 20 2.2.1 人工神经网络的原理 ......................................... 20 2.2.2 人工神经网络的基本结构 ..................................... 21 2.2.3 人工神经网络的主要特征 ..................................... 22 2.2.4 人工神经网络的分类 ......................................... 23IV 2.2.5 人工神经网络具备的功能 ..................................... 25 2.3 人工神经网络技术与植物配置的关系 ............................... 26 2.3.1 与植物配置特性的关系 ....................................... 26 2.3.2 与植物配置过程的关系 ....................................... 27 2.3.3 应用可行性分析 ............................................. 29 2.4 本章小结 ....................................................... 30 第三章 技术操作可行性研究........................................... 31 3.1 理想模型的结构 ................................................. 31 3.2 基础训练数据的获取 ............................................. 31 3.2.1 植物配置训练数据的来源 ..................................... 32 3.2.2 人工梳理待训练数据 ......................................... 33 3.2.3 提取数据的具体操作 ......................................... 34 3.3 人工神经网络模型的应用 ......................................... 42 3.3.1 影响因子与人工神经元的映射 ................................. 43 3.3.2 环境因子设置 ............................................... 43 3.3.3 植物配置方法设置 ........................................... 45 3.3.4 人工评价体系引入 ........................................... 47 3.3.5 机器学习的过程 ............................................. 49 3.4 输出结果与设计软件的对接 ....................................... 50 3.5 本章小结 ....................................................... 53 第四章 未来技术在可行性上的优化..................................... 54 4.1 完整的风景园林方案生成方法 ..................................... 54 4.1.1 气泡图标注 ................................................. 55 4.1.2 设计方案输出 ............................................... 56 4.2 多神经网络联合系统 ............................................. 57 4.2.1 需求分析网络 ............................................... 57 4.2.2 功能空间布局网络， ......................................... 58 4.2.3 环境分析网络 ............................................... 59 4.2.4 植物配置网络 ............................................... 59V 4.2.5 造价控制网络 ............................................... 59 4.3 云端设计技术展望 ............................................... 60 4.4 智能时代的景观师角色 ........................................... 63 4.5 本章小结 ....................................................... 64 结 论 ........................................................... 65 全文总结...........................................................65 未来展望..............