基于乡镇尺度Landsat8 OLI影像融合算法适应性研究

Study on the Suitability of Landsat8 OLI Image for Fusion Algorithms Based on the Township Scale

本文利用OIF因子选择乡镇尺度下Landsat8 OLI影像MS最优波段组合,在此基础上,研究OLI影像MS波段与PAN波段对6种融合算法:Brovey法、PCA法、Daubechies小波变换法、Coifet小波变换法、HIS与小波相结合的变换法、PCA与小波相结合的变换法融合的适应性,并对融合前后影像进行SVM分类,以验证融合结果在实际生产应用中的有效性。结果表明:B456为7波段35种组合方式中最佳波段组合,其OIF值为27.842;对融合前后影像进行定性和定量精度评价,OLI影像对PCA算法融合适应性最强,各精度指标均占优;Daubechies小波算法光谱扭曲度最小;HIS-wavelet算法清晰度最高;PCA-wavelet算法相关系数最高,融合结果信息含量最大;适应性最差为Brovey算法。土地利用分类精度验证结果表明:OLI影像经PCA算法融合后有助于提高分类精度。

In this study, we used the OIF factor to choose the best MS band combination for Landsat8 OLI image at the township scale, aiming to study the suitability of 6 kinds of fusion algorithms including standard color variation method (Brovey method), the principal component transformation method (PCA method), Daubechies transformation method for wavelet, Coifet wavelet transformation method, transformation method combining HIS with wavelet and PCA combining with wavelet transformation method for merging of MS and PAN brands of OLI images, and classified the image before and after merging to verify the validity of the fusion results in actual production application with SVM method. Results showed that the B456 brand which OIF value was 27.842 was the best band combination among 35 kinds of combinations of 7 bands. Qualitative and quantitative accuracy assessment before and after merging image showed that each index of OLI image was the dominated for PCA algorithm, which had the highest merging adaptation. And the spectral distortion degree of Daubechies wavelet algorithm was the smallest; HIS-wavelet algorithm had the highest sharpness; PCA-wavelet algorithm had the highest correlation coefficient and merging information contents compared with others. Brovey was the worst adaptive algorithm among 6 kinds of fusion algorithms;Accuracy verification of land use classification demonstrated that OLI image which merged by PCA algorithm would contribute to improve the classification accuracy.