遥感异常目标的仿生非线性滤波检测
发布时间：2019-06-15 17:40
【摘要】：目的为了解决复杂背景干扰下基于线性滤波异常检测算法无法有效区分复杂背景特征与异常目标特征,导致检测结果虚警率偏高的问题,提出一种面向复杂背景的遥感异常小目标仿生非线性滤波检测算法。方法受生物视觉系统利用不同属性信息挖掘高维特征机理的启发,该算法通过相关型非线性滤波器综合多波段光谱数据提取高维光谱变化特征作为异常目标检测检测依据,弥补线性滤波抗噪性能差,难于区分复杂背景特征与目标特征的缺点。结果仿真实验结果验证该算法在仿真数据及真实遥感数据的异常检测效果上有较大改善,在实现快速异常检测的同时提高了检测命中率。结论本文方法不涉及背景建模,计算复杂度低,具有较好的实时性与普适性。特别是对复杂背景下的小尺寸异常目标具有较好的检测效果。
[Abstract]:Aim in order to solve the problem that the anomaly detection algorithm based on linear filtering can not effectively distinguish the complex background features from the abnormal target features under complex background interference, and lead to the high false alarm rate of the detection results, a bionic nonlinear filtering algorithm for remote sensing abnormal small targets for complex background is proposed. Methods inspired by the mechanism of mining high-dimensional features by using different attribute information in biological vision system, the algorithm uses correlation nonlinear filter to synthesize multi-band spectral data to extract high-dimensional spectral variation features as the basis of abnormal target detection, which makes up for the poor anti-noise performance of linear filtering and difficult to distinguish complex background features from target features. Results the simulation results show that the algorithm improves the anomaly detection effect of simulation data and real remote sensing data, and improves the hit rate of detection while realizing fast anomaly detection. Conclusion the proposed method does not involve background modeling, has low computational complexity, and has good real-time and universality. Especially, it has a good detection effect for small size abnormal targets in complex background.
【作者单位】： 河海大学物联网工程学院;
【基金】：国家自然科学基金项目(41301448,61573128,61273170)~~
【分类号】：TP751
[Abstract]:Aim in order to solve the problem that the anomaly detection algorithm based on linear filtering can not effectively distinguish the complex background features from the abnormal target features under complex background interference, and lead to the high false alarm rate of the detection results, a bionic nonlinear filtering algorithm for remote sensing abnormal small targets for complex background is proposed. Methods inspired by the mechanism of mining high-dimensional features by using different attribute information in biological vision system, the algorithm uses correlation nonlinear filter to synthesize multi-band spectral data to extract high-dimensional spectral variation features as the basis of abnormal target detection, which makes up for the poor anti-noise performance of linear filtering and difficult to distinguish complex background features from target features. Results the simulation results show that the algorithm improves the anomaly detection effect of simulation data and real remote sensing data, and improves the hit rate of detection while realizing fast anomaly detection. Conclusion the proposed method does not involve background modeling, has low computational complexity, and has good real-time and universality. Especially, it has a good detection effect for small size abnormal targets in complex background.
【作者单位】： 河海大学物联网工程学院;
【基金】：国家自然科学基金项目(41301448,61573128,61273170)~~
【分类号】：TP751
【相似文献】
相关期刊论文 前10条
1 李俊生;图像非线性滤波技术的研究[J];常州工学院学报;2005年02期
2 鹿传国;冯新喜;孔云波;张迪;;几种非线性滤波只测角跟踪性能评估[J];弹箭与制导学报;2011年04期
3 冯驰;杜云明;;一种高效的非线性滤波技术[J];应用科技;2007年04期
4 赵静;黄晋英;王夙U，
本文编号：2500395
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2500395.html
