基于Cubist模型树的城市不透水面百分比遥感估算模型
发布时间：2019-02-19 11:51
【摘要】：不透水面是城市区域中一种典型的土地覆盖类型,是衡量城市环境质量和城市化水平的重要标志之一。与传统基于像元级的遥感研究方法相比,不透水面百分比(Impervious Surface Percent,ISP)的估算可以进入像元内部,获得更准确的城市信息。本文应用Cubist模型树,对Landsat TM的原始波段变量(除热红外波段),建立ISP估算的基础模型(Base Cubist-ISP)。通过基于模型树的集成学习优化算法和加入相邻时相影像的波段变量中值,以削弱噪声的影响。然后,优选热红外波段和各种衍生变量,并进行属性精简,继而应用集成学习算法得到的参数和精简后的变量建立ISP估算的优化模型(Optimal CubistISP)。对广东省广州市海珠区的实验结果表明,Optimal Cubist-ISP模型估算不透水面的整体均方根误差(RMSE)为12.98%,决定系数(R2)为0.90,精度明显优于Base Cubist-ISP模型,RMSE降低约5.03%,ISP在透水面区域被高估和高密度不透水面区域被低估的现象得到改善。本文提出的基于Cubist模型树建立ISP遥感估算的模型及优化方法可以适用于城市区ISP的提取。
[Abstract]:Impervious surface is a typical land cover type in urban area, which is one of the important indicators to measure the quality of urban environment and the level of urbanization. Compared with the traditional pixel-based remote sensing method, the estimation of impermeable surface percentage (Impervious Surface Percent,ISP) can enter into the pixel and obtain more accurate city information. In this paper, the basic model of ISP estimation (Base Cubist-ISP) is established for the original band variables of Landsat TM (excluding heat infrared band) by using the Cubist model tree. An ensemble learning optimization algorithm based on model tree and the median value of band variables of adjacent temporal images are used to reduce the influence of noise. Then, the thermal infrared band and various derived variables are selected optimally, and the attributes are reduced. Then the parameters obtained by the integrated learning algorithm and the reduced variables are applied to establish the optimization model of ISP estimation (Optimal CubistISP). The experimental results of Haizhu District, Guangzhou City, Guangdong Province show that the total root mean square error (RMSE) and determination coefficient (R2) of Optimal Cubist-ISP model are 12.98 and 0.90 respectively, which are better than that of Base Cubist-ISP model. The decrease in RMSE was about 5.03%, and the overestimation of ISPs in the pervious area and the underestimation of the high density impervious surface were improved. The model and optimization method of ISP remote sensing estimation based on Cubist model tree proposed in this paper can be applied to the extraction of ISP in urban areas.
【作者单位】： 华南师范大学地理科学学院广东省智慧国土工程技术研究中心;
【基金】：国家自然科学基金项目(41101152) “973”计划前期研究专项(2014CB460614) 广州市产学研协同创新重大专项民生科技项目(156100021) 广东省科技计划项目(2015A010103013)
【分类号】：P237
，
本文编号：2426474
[Abstract]:Impervious surface is a typical land cover type in urban area, which is one of the important indicators to measure the quality of urban environment and the level of urbanization. Compared with the traditional pixel-based remote sensing method, the estimation of impermeable surface percentage (Impervious Surface Percent,ISP) can enter into the pixel and obtain more accurate city information. In this paper, the basic model of ISP estimation (Base Cubist-ISP) is established for the original band variables of Landsat TM (excluding heat infrared band) by using the Cubist model tree. An ensemble learning optimization algorithm based on model tree and the median value of band variables of adjacent temporal images are used to reduce the influence of noise. Then, the thermal infrared band and various derived variables are selected optimally, and the attributes are reduced. Then the parameters obtained by the integrated learning algorithm and the reduced variables are applied to establish the optimization model of ISP estimation (Optimal CubistISP). The experimental results of Haizhu District, Guangzhou City, Guangdong Province show that the total root mean square error (RMSE) and determination coefficient (R2) of Optimal Cubist-ISP model are 12.98 and 0.90 respectively, which are better than that of Base Cubist-ISP model. The decrease in RMSE was about 5.03%, and the overestimation of ISPs in the pervious area and the underestimation of the high density impervious surface were improved. The model and optimization method of ISP remote sensing estimation based on Cubist model tree proposed in this paper can be applied to the extraction of ISP in urban areas.
【作者单位】： 华南师范大学地理科学学院广东省智慧国土工程技术研究中心;
【基金】：国家自然科学基金项目(41101152) “973”计划前期研究专项(2014CB460614) 广州市产学研协同创新重大专项民生科技项目(156100021) 广东省科技计划项目(2015A010103013)
【分类号】：P237
，
本文编号：2426474
本文链接：https://www.wllwen.com/kejilunwen/dizhicehuilunwen/2426474.html
