摘要
首先,针对研究区GF-2影像进行Brovey变换、G-S变换、NNDpansharp变换、PC变换4种融合,对融合结果进行定量评价;其次,利用随机森林分类方法对研究区作物进行分类,并进行精度验证,提出了研究区域农作物信息。结果表明:1)对研究区进行4种方法融合,提高遥感影像分辨率;2)从评价结果可知,4种融合影像中,NNDpansharp融合影像质量最佳。分类结果说明,NNDpansharp融合影像的随机森林分类总精度和Kappa系数最高,该方法和结果可为农业部门将高分二号遥感影像融合提取棉花面积方法提供选择性参考。
In order to obtain the most accurate cotton area data,four commonly used remote sensing image fusion methods,such as Brovey,G-S,NNDpansharp and PC are applied to full-color and multi-spectral band fusion for sub-meter GF-2 satellite data.Besides,visual interpretation based subjective evaluation and grey average,standard deviation,information entropy,average gradient,correlation coefficient and other quality parameters based objective evaluation methods are applicated to assess the four fusion results.Finally,the random forest classification method was used to classify the five types of land features such as cotton,corn,orchard,water body and other areas in the study area,and classification accuracy was calculated to identify the optimal fusion image.The calculation results showed that:1)Among four fusion methods,the NNDpansharp fusion quality was the best.2)Explanation of classification results,NNDpansharp fusion images have the highest total accuracy of random forest classification and Kappa coefficient.This method and results can provide a selective reference for the agricultural department to extract cotton area by fusion of GF-2 remote sensing images.
作者
阿依谢姆·米吉提
买买提·沙吾提
AYIXIEMU Mijiti;MAIMAITI Shawuti(Collge of Resources and Environmental Sciences,Xinjiang University,Urumqi 830046,China;Ministry of Education Key Laboratory of Oasis Ecology at Xinjiang University,Urumqi 830046,China;Xinjiang Agricultural Resources and Regional Planning Office,Urumqi 830046,China)
出处
《测绘与空间地理信息》
2023年第4期81-84,88,共5页
Geomatics & Spatial Information Technology
关键词
GF-2影像
融合图像
信息熵
随机森林分类
GF-2 images
image fusion
information entropy
random forest classification