基于互补特征的桥梁识别方法  

A BRIDGE RECOGNITION METHOD BASED ON COMPLEMENTARY FEATURES

作　　者：张永梅[1] 孙静[1] 叶晨[2] 

机构地区：[1]北方工业大学信息工程学院,北京100144 [2]中国石油信息技术服务中心,北京100007

出　　处：《计算机应用与软件》2014年第3期151-155,158,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61371143);国家科技支撑计划资助项目(2012BAH04F00);北京市自然基金项目(4132026)

摘　　要：针对单一图像进行桥梁识别的运行速度慢、误检率高的问题,提出一种结合多源遥感图像互补特征的桥梁识别方法。利用多光谱图像的颜色特征进行水陆分割,得到潜在桥梁区域,分别提取全色、SAR图像潜在桥梁区域的纹理特征和区域均值比特征,结合三者结果,得到桥梁识别结果。实验结果表明,结合多源遥感图像互补特征的桥梁识别方法有效提高了桥梁识别的准确率。For the problems of slower running speed and higher error rate in bridge recognition with a single image, we present a bridge recognition method which combines the complementary features of multi-source remote sensing images. We segment the land and water using the colour features of multi-spectral images and get potential bridge areas, then we extract respectively the texture feature and the regional mean ratio characteristics of the potential bridge areas from panchromatic and SAR images, and combine the results of these three to obtain the final recognition results. Experimental results show that the bridge recognition method combining complementary features of multi-source remote sensing images effectively improves the accuracy of bridge recognition.

关 键 词：互补特征 目标识别 多源图像 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]