| 摘要: |
| 结合大数据的街景，风景园林师能够快速、有效地对建成环境的感受进行评价。但是目前街景分析主要依靠图像内容语义分割等高级认知，较少考虑人视知觉初级加工特征。研究证据表明在视知觉初级加工中，丰富的高频细节信息可以提升人们对环境的喜好并强化环境恢复效果。为了探索不同景观风貌的细节差异，筛选了150张自然景观、69张城市景观、60张不同林冠覆盖率的树林景观以及31张中国古典园林典型景观图像，并对代表它们的高频细节信息进行了傅里叶斜率分析。结果表明，看似简单的自然要素，如岩体和砂石、水面、修剪草坪等，比城市人工景观的细节要丰富多变，其傅里叶斜率仍然更接近自然。特别指出的是，中国古典园林的傅里叶斜率更加接近典型自然，呈现出更加丰富、更加类似于自然的复杂细节。本研究拓展了提升街道环境感受的可能的途径，即除了利用植物等自然要素之外，还有可能通过模拟自然丰富多变的细节和形态来实现。 |
| 关键词: 景观风貌 傅里叶斜率 幂律特征 数字图像处理 分形 |
| DOI：10.14085/j.fjyl.2021.12.0085.08 |
| 分类号:TU986 |
| 基金项目:国家自然科学基金（编号 51878461） |
|
| Landscape in Details: High-Frequency Details and Power-Law Characteristics of Landscape Images |
|
CHEN Zheng, LI Ye
|
|
Tongji University
|
| Abstract: |
| Street view analysis backed with big data can help landscape architects quickly and effectively evaluate the feelings of the built environment. However, the current street view analysis mainly relies on advanced cognition such as semantic segmentation of image content, which seldom considers the characteristics of the primary processing of human visual perception. Research evidences show that in the primary processing of visual perception, rich high-frequency detailed information can enhance people's preference for the environment and strengthen its recovery effect. To explore the differences in the details of different landscape images, the authors select typical landscape images, including 150 natural landscapes, 69 urban landscapes, 60 forest landscapes with diversified canopy coverage, and 31 Chinese classical garden landscapes, to analyze the Fourier slopes of their high-frequency details. The results show that the seemingly simple natural elements, such as rock mass, sand, water surface, and lawn pruning, are richer and more varied than the details of urban artificial landscapes, and their corresponding Fourier slope is still closer to nature, and the tree canopy details are higher than the typical natural landscapes. In particular, the Fourier slope of Chinese classical gardens is closer to typical nature, presenting richer, more nature-like complex details. This research expands the possible ways to improve the street environment experience. In addition to utilizing natural elements such as plants, simulating the rich and varied details and forms of nature can have the same effect. |
| Key words: landscape Fourier slope power-law characteristics digital image processing fractal dimension |
