摘要
为了实现快速、自动化发现土地覆盖变化这一目标,在分析传统主成分差异法、差异主成分法、多波段主成分法三种不同处理过程的基础上,结合主成分变换原理提出了一种改进的主成分分析法(modified principal component analysis,MPCA)。操作中先将d1时相多光谱影像作主成分分析,得PC1d1,PC2d1,…,PC6d1;d2时相高分辨率全色波段PAN与PC1d1进行直方图匹配后,采用了经反复试验效果较好的3×3模板进行边缘滤波增强;然后取代PC1d1与PC2d1,PC3d1,…,PC6d1进行主成分逆变换,作者在ENVI4.0和IDL6.0工具包支持下实现了这一融合算法。以北京海淀区为例进行的试验研究表明,MPCA法不仅能够快速发现变化信息,而且增强了影像纹理,弥补了传统主成分分析法的缺陷。此外,变化信息提取精度较高,其Kappa系数比传统主成分差异法、差异主成分法、多波段主成分法依次提高了0.063,0.118,0.029,是一种比较实用的变化信息发现方法,值得推广应用。
To achieve the goal of quickly and automatically detecting land cover change, on the basis of analyzing traditional PCA differentia, differentia PCA and multi-band principal component transformation, combined with the principle of the principal component transformation, a modified principal component analysis （MPCA） is presented and implemented based on ENVI 4.0 and IDL 6.0 saddlebag. The operation is as follows： firstly, using PCA method for multi-spectral images in d1 time phase; secondly, making histogram matching between high resolution PAN band in d2 time phase and the first principle component in d1 time phase（PC1d1 ）, then putting up edge filter enhancement with 3 × 3 template tested fruitfully time after time; thirdly, replacing PC1d1 by PAN enhanced and carrying through inverse principal components with other principal components in d1 time phase. In this paper, taking Haidian district in Beijing for example, different PCA methods have been tried out. It is found that the presented MPCA method not only can quickly detect change information, but also has improved the texture of image and made up the defect of the traditional PCA smoothing texture. Furthermore, the precision of MPCA extracting change information is high, and its Kappa coefficients are 0. 063, 0. 118, 0. 029, higher than that of the traditional PCA differentia, differentia PCA and multi-band principal component transformation, respectively. The result shows that MPCA method is a more practical method for detecting land cover change and deserves popularization and application.
出处
《成都理工大学学报（自然科学版）》
CAS
CSCD
北大核心
2007年第1期92-96,共5页
Journal of Chengdu University of Technology: Science & Technology Edition
基金
四川省青年科技基金(06ZQ026-014)
四川省教育厅自然科学重点项目(2006A116)
