| 引用本文： |
-
郭茂祖,王偲佳,王鹏跃,李阳,赵玲玲.基于过采样和卫星图像多特征融合的街区品质评估[J].广西科学,2022,29(4):652-659. [点击复制]
- GUO Maozu,WANG Sijia,WANG Pengyue,LI Yang,ZHAO Lingling.Block Quality Evaluation Based on Oversampling and Satellite Image Multi-feature Fusion[J].Guangxi Sciences,2022,29(4):652-659. [点击复制]
|
|
| 本文已被：浏览 258次 下载 256次 |
码上扫一扫！ |
| 基于过采样和卫星图像多特征融合的街区品质评估 |
|
郭茂祖1,2, 王偲佳1,2, 王鹏跃2,3, 李阳1,2, 赵玲玲4
|
|
|
| (1.北京建筑大学电气与信息工程学院, 北京 100044;2.建筑大数据智能处理方法研究北京重点实验室(北京建筑大学), 北京 100044;3.北京建筑大学建筑与城市规划学院, 北京 100044;4.哈尔滨工业大学计算机科学与技术学院, 黑龙江哈尔滨 150001) |
|
| 摘要: |
| 城市街区的品质评估是城市规划中的重要步骤之一。针对已有街区品质评估模型中使用的街景图像采集位置和角度不统一造成的图像内容偏差问题，本文提出使用视角统一且易获取的卫星图像代替街景图像。同时考虑多种街区品质测度指标，将非图像型数据与图像型数据进行特征融合，从多角度进行城市街区品质表达，然后针对小样本和数据不平衡场景，结合机器学习中的随机过采样和随机森林（Random Forest，RF），基于融合数据进行城市街区的品质评估。实验结果表明，本研究方法可以有效进行全面的城市街区品质表达，并有效降低过拟合，对比未使用过采样和多特征融合的方法，正确率提高约8个百分点。 |
| 关键词: 城市街区品质评估 特征融合 朴素随机过采样 随机森林 卫星图像 |
| DOI：10.13656/j.cnki.gxkx.20220919.005 |
| 投稿时间：2022-04-30 |
| 基金项目:国家自然科学基金面上项目（61871020，62101022），北京建筑大学市属高校基本科研业务费专项（X20161）和北京市属高校高水平创新团队建设计划项目（IDHT20190506）资助。 |
|
| Block Quality Evaluation Based on Oversampling and Satellite Image Multi-feature Fusion |
|
GUO Maozu1,2, WANG Sijia1,2, WANG Pengyue2,3, LI Yang1,2, ZHAO Lingling4
|
| (1.College of Electrical and Information Engineering, Beijing University of Civil Engineering and Architecture, Beijing, 100044, China;2.Beijing Key Laboratory of Intelligent Processing for Building Big Data (Beijing University of Civil Engineering and Architecture), Beijing, 100044, China;3.College of Architecture and Urban Planning, Beijing University of Civil Engineering and Architecture, Beijing, 100044, China;4.College of Computer Science and Technology, Harbin Institute of Technology, Harbin, Heilongjiang, 150001, China) |
| Abstract: |
| The quality evaluation of urban blocks is one of the important steps in urban planning.In order to solve the problem of image content deviation caused by the inconsistent collection position and angle of street view image used in the existing block quality evaluation model, this article proposes to use the satellite image with unified perspective and easy access to replace the street view image.At the same time, considering a variety of block quality measurement indicators, the characteristics of non image data and image data are fused, and the quality of urban blocks is expressed from multiple perspectives.Then, for small samples and data imbalance scenarios, combined with random oversampling and Random Forest (RF) in machine learning, the quality of urban blocks is evaluated based on the fusion data.The experimental results show that this research method can effectively carry out a comprehensive urban block quality expression and effectively reduce over fitting.Compared with the methods without over sampling and multi-feature fusion, the accuracy rate is increased by about 8 percentage points. |
| Key words: quality evaluation of urban blocks feature fusion simple random oversampling random forest satellite image |
