一种针对多光谱遥感图像的桥梁快速识别算法

提出了一种针对多光谱图像中桥梁的识别算法。首先,根据水体和背景地物在不同光谱波段的亮度差异,计算多光谱图像的水体指数得到水体增强图,搜索其具有明显双峰的直方图得到最优阈值,实现河流的完整提取；其次,利用桥梁的存在会导致局部水体的光谱异常,沿河流中间线进行潜在桥梁区域的快速提取；再进一步利用桥梁长度以及与河流的空间关系进行鉴别,有效剔除虚警。利用 SPOT4遥感影像进行实验,结果表明本文算法运算量小,对于多个桥梁的识别具有很好的实用性。

A Fast Bridge Recognition Algorithm for Multi-Spectral Remote-Sensing Image

A fast bridge recognition algorithm for multi-spectral remote sensing image was proposed. Firstly, according tothe brightness difference between water and background object in different spectral band, the water index was computed toform the water enhanced image. The optimal threshold could be found by searching the enhanced histogram which had ob.vious two peaks. Since the present of bridge would result in spectral abnormal in local water regions, a detector was used toextract the potential bridge regions along river centerline. Then the length of bridge and the spatial relationship was usedto discard possible false alarms. Experiments were executed on SPOT4 images and the results showed that the method is ef.fective and practical for bridge detection.