【摘要】 以60年代宁波平原地区的锁眼卫星为数据源,提出一种基于PS图像处理方法来实现河网的提取,并与阈值法、监督分类法和决策树法进行对比分析。结果表明,传统的阈值法和监督分类方法对于单波段高分辨率遥感影像的提取能力十分有限,决策树方法和基于PS图像处理法取得的面状数据整体精度在70%以上,线状数据整体精度在60%以上。在提取水面信息的能力上,决策树方法的正确率较高,但图斑较为破碎,水网连接度保留不完整;基于PS图像处理法受建筑阴影干扰较大,但水体提取完整度较高,水网连接度保留效果较好。更多还原

【Abstract】 In this paper, taking keyhole satellite data of Ningbo plain in the 1960 s as the data source, we proposed a new hydrology network extraction method based on the PS image processing method. Then, we compared this method with the traditional classi cation methods, such as threshold method, supervised classi cation method and decision tree method. e results demonstrate that the threshold method and supervised classi cation method have very limited ability to extract hydrology network information from single-band high-resolution remote sensing images. e overall accuracy of the decision tree method and the PS image processing method on hydrology surface information and line information is over 70% and 60% respectively. The extraction correctness of decision tree method is high, but the patch is fragmented and the connection of hydrology network is badly reserved. e PS image processing method shows poor ability to suppress interference of architectural shadows, but the extraction completeness is higher, and the hydrology network connectivity retention is better.更多还原