高分辨率遥感图像中桥梁自动识别方法研究  

Research on Methods of Automatic Bridge Recognition in High-resolution Remote Sensing Images

作　　者：杨浩[1] 尹东[1] 洪日昌[1] 

机构地区：[1]中国科学技术大学电子工程与信息科学系,安徽合肥230026

出　　处：《计算机仿真》2006年第9期119-122,共4页Computer Simulation

基　　金：国家"863"项目(2002AA783055)

摘　　要：提出一种针对高分辨率遥感图像中桥梁自动识别新算法。该算法从桥梁与河流的位置关系出发,结合遥感图像的成像特点和桥梁目标的形状特征,运用模糊处理,轮廓搜索和直线拟合识别桥梁目标。整个算法流程大致可分为三个步骤,首先在底层处理运用二维特征模糊分割水体区域;然后在中层处理中,通过水体轮廓搜索潜在桥梁点,并进行直线拟合提取初略桥梁目标;后处理中则对由于环境干扰所造成的断点进行连接,恢复出完整桥梁。试验结果表明该文算法具有识别速度快,识别准确率高及虚警率低等优点,且整个算法完全自动,不需人工干预。A novel algorithm for recognizing bridges automatically in the high - resolution remote sensing images is presented in this paper. Starting from the space relativity between the bridge and the river, and combining it with the imaging characteristics of remote sensing images and shape features of the bridge, the algorithm integrates fuzzy processing, contour searching and line fitting to recognize bridges. The whole algorithm is made up of three steps ： in the pre - processing, 2 - dimensional fuzzy algorithm is utilized to segment water areas; in the mid - level processing, the contours of water areas are extracted to search potential bridge pixels, and a line - fitting method is applied to extract candidate bridges ; in the post - processing, broken bridges due to the environmental interference are connected to restore the whole bridge. The experiment results show the good qualities of the algorithm, such as fast speed, high rate of recognition and low rate of false target. Moreover, the whole algorithm is entirely automatic, it does not need any manual interference.

关 键 词：遥感图像 自动识别 模糊处理 轮廓搜索 直线拟合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]