基于视觉注意机制和Hough变换融合的遥感影像油罐提取  

Oil Tank Extraction from Remote Sensing Images Based on Visual Attention Mechanism and Hough Transform

作　　者：吴晓东[1,2] 冯伍法[1] 冯倩倩 李润生[1] 赵爽 

机构地区：[1]信息工程大学 [2]61206部队 [3]河南省数联测绘科技有限公司

出　　处：《信息工程大学学报》2015年第4期503-506,共4页Journal of Information Engineering University

摘　　要：Hough变换用于遥感影像地物提取时具有较高的精度,但其算法非常复杂,效率较低。为提高Hough变换处理效率,提出了一种融合视觉注意机制和Hough变换遥感影像提取算法,并以油罐提取为例验证其有效性。首先提取影像的视觉显著性特征图,获取油罐目标显著性区域;然后对显著性区域进行Hough变换检测,得到油罐目标。实验结果证明,该方法能大幅减少Hough变换的计算量,提高效率的同时,提取精度也较高。Hough transform can achieve high precision in remote sensing image feature extraction, but the algorithm is very complex and inefficient. To improve the Hough transform processing effi- ciency, this paper presents a feature extraction algorithm for remote sensing images combined with visual attention mechanism and Hough transform. Test of oil tanks extraction are made to verify its validity as an example. Firstly, visual salient feature maps are extracted to get significant area of oil tank target. Then, Hough transform detection in significant area is implemented to get the tank target. The results show that the method can significantly reduce the calculation amount of Hough transform and improve efficiency and extraction accuracy.

关 键 词：视觉注意机制 视觉显著性 油罐提取 HOUGH变换 

分 类 号：P237[天文地球—摄影测量与遥感]