遥感图像中的黄河主溜线高效提取方法

Effective extraction method of the Yellow River mainstream from remote sensing image

黄河主溜线的趋向直接影响黄河的防洪决策与规划,而现有的主溜线演进方法需判断每一步河流方向。提出一种河流直向化的预处理方法,利用光谱相似性原理提取河流域,然后轮廓跟踪河流域外边界,对边界内的河流域通过线性插值的方式进行直向化处理;动态演进直向河流的主溜线,随后将其逆映射回原图像,获得原图像中河流域的主溜线。结果表明,运用直向化预处理方法提取主溜线,在没有降低误差精度的基础上,避免了频繁地检测河流走向,简化了主溜线的演进模型。

The tendency of the Yellow River’s mainstream line directly affects the flood prevention decision-making and planning. The existing method of evolving mainstream line needs to calculate the direction of the river at every step. A straight pretreatment method is proposed in this paper. First, it uses spectral similarity principle to extract the region of the river, then exerts contour tracking algorithm to get external boundary of the river, and finally utilizes linear interpolation to transform the river image in curve external boundary to a straight image. Next, it evolves the mainstream line dynamically in straight image and inversely maps to original image. The results of the experiment show that this method can avoid detecting the river direction frequently and simplify the evolution model of mainstream line without reducing the error precision.