中文摘要：

在系统归纳和分析现有的路口遥感信息提取方法的基础上,提出一种面向高空间分辨率遥感影像的路口自动定位新方法。该方法首先通过低梯度运算获取同质区域;然后设定阈值去除同质区内的水体、阴影以及小面元干扰物;再利用Hough变换检测二值图像中的直线,并根据直线参数出现的频率排序,保留参数出现频率较高且相互间夹角较大的直线;最后用该组直线交点的平均值定位路口。以福州市城区局部QuickBird全色影像为数据源定位四岔路口与三岔路口的实证研究表明,在同物异谱与异物同谱现象严重情况下,本文算法所定位的路口仍然准确有效。

英文摘要：

On the basis of systematically inducing and analyzing the methods to extract road junctions from remotely sensed imagery,a new approach to automatically positioning road junctions in high spatial resolution images is presented in this paper.The related algorithm includes the following steps： firstly,the homogeneous areas are obtained by lower gradient operator;then,waters,shadows and small areal distracting features are removed one by one;after that,the straight lines are detected in the binary image by means of Hough transform;finally,the detected straight lines are sorted according to the frequency with which they appear,and the road junctions are indicated with the average coordinates of the intersections of several top frequent straight lines each of which at least has one intersection angle larger than the predetermined threshold.In the case study,an intersection and a junction of three roads were located in local urban area of Fuzhou by using QuickBird panchromatic image as sample data,showing that the proposed approach,with robustness against spectral confusion and confused features,can efficiently and accurately locate road junctions.