摘要
分类器具有较强的信息挖掘能力,已被广泛应用于变化检测。多分类器集成可以综合利用不同分类器的优势以提取变化信息,因此,本文利用异质性较大的两类分类器——极限学习机(ELM)和支持向量机(SVM)构建集成结构。通常像素级变化检测可以较好地保存边缘信息,对象级变化检测可以较好地抑制噪声,两者可相互补充。本文利用较小异质度的多尺度分割对象对初始像素级变化检测结果进行约束,根据判定准则得到最终的变化检测结果。该方法既消除了像素级变化检测方法中普遍存在的"椒盐噪声",又减少了面向对象中分割尺度对变化检测的影响。通过徐州市的两景资源三号(ZY-3)遥感影像进行实验,对城市典型地物要素进行变化检测研究,证明本文构建的算法可以综合利用基于像素和面向对象两种方法的优势,能够有效提高变化检测方法的精度和稳定性。
The classifier is widely used in change detection due to its strong information mining ability.And the multiple classifier ensemble system can take advantages of different classifiers to extract change information,so in this paper two kinds of base classifiers （extreme learning machine （ELM） and support vector machine （SVM）） with the highest diversity are utilized to construct the ensemble system.Because of the limitation and disadvantages of the pixel-wise change detection and the object-oriented change detection methods for high-resolution remote sensing images,the initial pixel wise change map was combined with the smaller heterogeneity multi-scale segmentation map to obtain the final change map.The proposed method can reduce the commission ration and the impact of image segmentation on change detection.The experiment was carried out by two sets of ZY-3 remote sensing images to discuss the changes of typical urban features in Xuzhou,Jiangsu Province.In order to make full use of the rich spatial information of high-spatial-resolution remote sensing images,the texture and morphological features were extracted to construct the multi-source features along with spectral feature.The experimental results show that the proposed algorithm can integrate the advantages of both pixel-wise and object-oriented methods,and can effectively improve the accuracy and stability of the change detection method.
作者
张玉沙
黄岩
谭琨
陈宇
杜培军
ZHANG Yu-sha;HUANG Yan;TAN Kun;CHEN Yu;DU Pei-jun(School of Environment Science and Spatial Informatics,China University of Mining and Technology,Xuzhou 221116;Geological Exploration Technology Institute of Jiangsu Province,Nanjing 210049;School of Geographic and Oceanographic Sciences,Nanjing University,Nanjing 210023,China)
出处
《地理与地理信息科学》
CSCD
北大核心
2018年第3期54-60,共7页
Geography and Geo-Information Science
基金
国家自然科学基金项目(41471356)
徐州科技基金项目(KC16SS092)
江苏省优势学科资助项目
关键词
多分类器集成
多特征融合
异质性
影像分割
multiple classifier
multi-feature fusion
diversity
image segmentation
