摘要:
目前不同中空间分辨率遥感卫星在利用混合像元分解方法提取石漠化信息上存在效果差异,比较不同遥感卫星在提取石漠化信息上的差异,有助于进一步提高石漠化信息提取精度。本研究以贵州省普定县为例,采用GF-6与Landsat8卫星数据,利用顶点成分分析(vertex component analysis,VCA)和完全约束最小二乘法(fully constrained least squares,FCLS)相结合的混合像元分解方法进行石漠化信息提取,探究GF-6与Landsat8在石漠化信息提取的端元特征和等级差异,以此探索GF-6在提取石漠化信息的可行性与有效性。研究结果表明: ①红边波段范围上,GF-6植被端元波谱曲线明显区别于基岩与土壤端元,更易识别出植被端元; ②石漠化信息端元提取精度上,GF-6和Landsat8提取植被端元OA分别为0.63和0.45,Kappa系数分别为0.50和0.29,RMSE分别为1.19和1.71,GF-6和Landsat8提取基岩端元OA分别为0.79和0.61,Kappa系数分别为0.63和0.42,RMSE分别为0.54和0.88; ③石漠化等级评价上,GF-6和Landsat8提取石漠化等级OA分别为0.76和0.59,Kappa系数分别为0.56和0.38,RMSE分别为0.64和1.27。因此,GF-6在混合像元分解提取石漠化信息精度要优于Landsat8,且GF-6的红边波段能更好地识别石漠化区域植被信息,基于GF-6的混合像元分解方法可作为一种石漠化监测手段应用于实际工作中。
Abstract:
Different moderate-resolution remote sensing satellites exhibit various effects in extracting rocky desertification information using the pixel unmixing method. Comparing these various effects can help further improve the extraction accuracy of rocky desertification information. By extracting information on rocky desertification in Puding County, Guizhou Province from GF-6 and Landsat8 using the pixel unmixing method, this study investigated the end member characteristics and desertification grade differences between GF-6 and Landsat8. Furthermore, this study explored the feasibility and effectiveness of GF-6 in extracting rocky desertification information. The results are as follows: ① Within the red-edge band of GF-6 data, the vegetation end member exhibited significantly different spectrum curves from bedrock and soil end members, making it easier to identify the vegetation end member. ② In terms of end member extraction accuracy of rocky desertification information, GF-6 and Landsat8 yielded overall accuracy (OA) of 0.63 and 0.45 in extracting the vegetation end member, respectively, corresponding to Kappa coefficients of 0.50 and 0.29 and RMSEs of 1.19 and 1.71, respectively. Moreover, GF-6 and Landsat8 yielded OA of 0.79 and 0.61 in extracting the bedrock end member, respectively, corresponding to Kappa coefficients of 0.63 and 0.42 and RMSEs of 0.54 and 0.88, respectively. ③ In the evaluation of rocky desertification grades, GF-6 and Landsat8 yielded OA of 0.76 and 0.59 in extracting rocky desertification grades, respectively, corresponding to Kappa coefficients of 0.56 and 0.38 and RMSEs of 0.64 and 1.27. Therefore, GF-6 outperforms Landsat8 in the accuracy of extracting rocky desertification information using the pixel unmixing method. In addition, the red-edge band of GF-6 data can effectively identify the vegetation information in areas with rocky desertification. In summary, the pixel unmixing method based on GF-6 data can be practically applied to rocky desertification monitoring.
下载: