基于Gabor滤波器的航空图像居民区域提取

摘要

Automatic extraction of resident regions from medium and low revolution aerial images plays a crucial role in geographic information system（GIS） updating and unmanned aerial vehicle（UAV） navigation.The parameters of Gabor filter in detail were analyzed,and a fast resident region extraction algorithm based on Gabor filter was proposed.The method composed of four steps.Firstly,the texture features of input image were extracted using Gabor filter.And then the texture features were smoothed by the kernel density estimation method to get the confidence image.Subsequently,the confidence image was segmented by the basic global threshold algorithm to acquire candidate regions.Ultimately,the candidate regions were verified by the geometric structure information.The method has low computational complexity,and can run within 0.42s in average.The efficiency and accuracy of the proposed algorithm were demonstrated by the experimental results.%中低分辨率航空图像中居民区域的自动提取对地理信息系统更新和无人机导航具有重要作用.详细分析了Gabor滤波器参数对纹理提取的影响,提出了一种基于Gabor滤波器的居民区快速提取算法.算法包含4步：运用Gabor滤波器分析图像纹理,采用核密度估计生成居民区域置信图像,进而计算自适应阈值分割置信图得到候选区域,最后根据区域几何形状去除干扰得到居民区.算法平均运算时间为0.42 s,实验结果表明了算法的高效性和准确性.