光子学报, 2014, 43 (9): 0910001, 网络出版: 2014-10-23
基于稀疏表示的可见光遥感图像飞机检测算法
Airplane Detection in Optical Remote Sensing Image Based on Sparse-Representation
模式识别 计算机视觉 遥感图像 飞机检测 刚性目标 轮廓稀疏表示 几何部件 部件模型 Pattern recognition Computer vision Remote sensing image Aircraft detection Rigid object Sparse representation of profile Geometric parts Part-based model
摘要
为解决当前遥感图像飞机检测方法在复杂背景下准确率低, 实现旋转不变困难的问题, 结合图像稀疏表示原理, 提出一种基于稀疏表示的飞机检测算法.该算法首先利用飞机是刚性目标且具有明显几何外观的特点, 构建飞机几何原子库; 然后建立飞机轮廓几何逼近的最优化方程, 在稀疏表示原理框架下, 得到飞机轮廓最优的几何部件组合; 最后, 以星型结构的部件模型为框架, 生成待检测图像的目标显著图并根据显著图定位出飞机.实验结果表明, 稀疏表示方法能自适应选取飞机部件, 部件数目较少且不易受光照、颜色和复杂背景的影响.与现有算法相比, 本文算法准确率达90%以上, 检测速度有较大的提高.
Abstract
In this paper, an airplane detection algorithm based on sparse representation is proposed to solve the problems of low detection precision in complicated backgrounds and difficult achivement of rotation invariant.Three steps are included in this algorithm:first, a geometrical dictionary is built according to the rigid property and the typical geometric appearance of the airplane to be detected;second, the optimal combination of the geometrical parts is obtained by solving the airplane profile approximation model set up under the framework of sparse representation theory;third, the object salient map is generated based on the star-structure part-based model and the location of the objects can be obtained from the salient map.Experimental results indicate that this algorithm can adaptively select the geometrical parts of airplane as few as possible and is insusceptible to illumination, color or complicated backgrounds.Compared to existing methods, the detection precision of our algorithm reaches above 90%, and the detection speed is significantly improved.
林煜东, 和红杰, 尹忠科, 陈帆. 基于稀疏表示的可见光遥感图像飞机检测算法[J]. 光子学报, 2014, 43(9): 0910001. LIN Yu-dong, HE Hong-jie, YIN Zhong-ke, CHEN Fan. Airplane Detection in Optical Remote Sensing Image Based on Sparse-Representation[J]. ACTA PHOTONICA SINICA, 2014, 43(9): 0910001.