基于国产GF-1遥感影像的山区细小水体提取方法研究

Extraction of small river information based on China-made GF-1 remote sense images

以国产"高分一号"16m遥感图像为数据源,在特克斯县选取两个研究区域,针对山区细小线状河流提取难度较大的问题,使用基于规则的面向对象的方法实现了对山区细小水体的精确化提取。首先,在总结前人选择最优分割尺度的基础上,考虑了各层权重信息,针对某一特定地物,提出了指示最优分割尺度的指标——改进的与邻域绝对均值差分方差比(MRMAS),并由此获取了影像上细小水体的最优分割尺度。其次,为区分水体和山体阴影,构建阴影水体指数SWI=B1+B2-B4,成功剔除了绝大部分阴影信息。最后,利用形态学膨胀滤波及Pavlidis异步细化算法对提取的细小水体进行后处理,最终得到细小河流的矢量化水系图。实验结果表明,该方法可以完整、快速地提取出山区细小线状河流信息,总体精度在90%以上,Kappa系数在85%以上,有效排除阴影等暗色地物的干扰,基本消除椒盐噪声。该研究成果或对国产高分影像处理系统的研发与应用提供一定的科学参考。

The object-oriented method was used to extract small water bodies in mountainous areas and resolve the difficult problem of interpreting small linear rivers in hilly areas. We did this using China- made GF- 1 remote sensed images with a 16 m × 16 m resolution. On the basis of previous studies we put forward an index considering the weight of each layer- Modified Ratio of Mean Difference to Neighbors(ABS)to Standard Deviation(MRMAS),which mainly reflects optimal segmentation scale for a particular object. Then the optimal segmentation scales of novel water bodies were obtained. The Nir threshold was used to get dark features,according to the characteristics of the water body objects including spectrum,shape,and spatial structure. The rectified shadow water index SWI=B1+B2-B4(B1:blue,B2:green,B4:near-infrared)was employed to distinguish between water and mountain shadow,which successfully eliminated most shadow information. Afterwards,small water information was obtained using DEM and Length/Width thresholds to remove a little amount of residual shadow. Last,we applied morphological dilation filtering and the Pavlidis asynchronous thinning algorithm to reprocess extracted tiny water to obtain small river vectorization. The proposed methodology is able to integrally extract small water bodies in hilly areas with overall accuracy of more than 90% and a Kappa coefficient more than85%. This methodology not only effectively eliminates the interference of dark surface features such as shadows,but largely eliminates the salt- phenomenon. This study provides scientific information for the development and application of the China- made GF- 1 remote sensing image processing system.