基于残余谱方法的农村居民地遥感提取研究

Study on Remote Sensing Extraction of Rural Residentjal Region Based on Spectral Residual Method

视觉注意模型通过模拟人类的视觉注意机制而得到图像中最容易引起注意的显著区域.残余谱(spectral residuals,SR)方法是视觉注意模型中的一种,通过分析自然图片在频率域中的log谱来检测图片的显著性目标.将残余谱方法引入到遥感影像中用于获取农村居民地信息,通过比较高分1号、资源3号和WorldView 2这3个不同传感器的卫星影像,试分析遥感影像中不同的波谱和影像不同的获取时间对居民地提取结果的影响.结果表明:卫星影像中RGB组合提取的居民地效果更好,近红外波段在残余谱方法提取居民地时具有消极影响;8月获取的卫星影像用于居民地的提取相比于4月效果更好;残余谱方法用于农村居民地的提取对传感器的选择无明显倾向性.

Visual attention model can get the most noticeable region from image by simulating human visual attention mechanism.Spectral residual(SR)method is one of the visual attention models.The salient targets of pictures are detected by cnalyzing the log spectrum of natural pictures in frequency domain.In this paper,spectral residual method lis employed to remote sensing imagery to obtain rural residential re-gions.By comparing satellite images of GF-1,ZY-3 and WorldView-2 these three different sensors,the effects of different spatial resolution and different acquisition times of satellite images on extracted results of residential regions are analyzed.The results show that the effect of residential regions extracted by RGB combination in satellite image is better,and near-infrared band has negative influence on spectral residual method when extracting residential regions.The satellite images acquired in August are better for residential regions extraction than those in April.The spectral residual method used for the extraction of rural residential re-gions has no obvious tendency to the selection of sensors in this experiment.