基于小波支持向量机分割的SAR图像桥梁目标检测

针对SAR图像中桥梁和水域的统计特性,提出了基于小波支持向量机分割与先验知识相结合的桥梁目标检测方法.通过对SAR图像中桥梁和背景的分析,首先对水域进行特征提取,再利用小波支持向量机方法对数据进行训练建模,通过模型对SAR图像中的河流进行分割,最后在分类后的二值图中按方向累加能量最小准则进行桥梁目标检测.基于真实SAR图像的实验结果显示,此方法不需对SAR图像进行复杂的预处理,有强的抗斑点噪声性,能快速、准确地检测SAR图像中的桥梁目标.

Detection of bridges based on WSVM segmenting in SAR image

After existing bridges and river regions statistical feature in the SAR images were considered, a novel method was proposed to detect bridges by combined wavelet support machine (WSVM) with knowledge of bridges. The feature of river domain was extracted by analyzing bridges and their backgrounds in SAR image. WSVM was used to make classification model by training example data for segmenting river region. The last direction energy function was used as the rule for identifying a bridge in the binary image of river class. The computer experiments were carried out to detect bridges from real SAR images. The proposed method is feasible without complex preprocessing under noise distributions. The experimental results showed that the method is fast and accurate for detecting bridges.