基于横纵多尺度灰度差异加权双边滤波的弱小目标检测
投稿时间:2019-08-21  修订日期:2019-10-23  点此下载全文
引用本文:
摘要点击次数: 74
全文下载次数: 0
作者单位E-mail
朱含露 中科院上海技术物理研究所 zhuhanlu19920@163.com 
张旭中 中国科学院湖州应用技术研究与产业化中心  
陈忻 中科院上海技术物理研究所  
胡亭亮 中科院上海技术物理研究所  
饶鹏 中科院上海技术物理研究所  
基金项目:光学技术与仪器项目
中文摘要:为了有效地检测复杂背景下的红外弱小目标,提出了一种基于横纵多尺度灰度差(HV-MSGD)的方法来增强弱目标,并通过距离和像素差异来实现对背景强边的抑制。目标区域与周围区域之间存在不连续性,为了加强它们的差异,HV-MSGD与双边滤波(BF)相结合,可以在抑制背景的同时提高目标强度。进一步通过自适应局部阈值分割和全局阈值分割来提取候选目标。为了进一步验证对单帧检测的影响,将上述单帧检测算法与改进的无迹卡尔曼粒子滤波器(UPF)相结合,实现轨迹检测。实验结果表明,该方法在弱信噪比(SNR)下优于其他方法,在抑制背景的同时可以增强目标,增强效果是其他方法的6-30倍。在实验中,输入信噪比分别为2.78,1.77,1.79,1.13和1.16。图像处理后,背景抑制因子(BSF)分别为13.48,21.33,11.73,20.63和121.92,信噪比增益(GSNRs)分别为40.09,71.37,27.53,12.65和131。该方法的检测概率(Pd)也优于其他算法。当误报率(FAR)为5×10^(-4), 1×10^(-3), 1×10^(-3), 1×10^(-5)和7×10^(-6),计算五组真实序列图像的Pd为94.4%,92.2%,91.3%,95.6%和96.7%。
中文关键词:弱目标检测  多尺度灰度差  距离和像素差  局部阈值分割  全局阈值分割
 
Dim Small Targets Detection based on Horizontal-Vertical Multi-scale Grayscale Difference weighted Bilateral Filtering
Abstract:For the efficient detection of infrared dim targets in complex background, a method based on horizontal-vertical multi-scale grayscale difference (HV-MSGD) is proposed to enhance the weak target, and the suppression of background strong edge is realized by combining distance and pixel difference. It is discontinuity between the target area and the surrounding area, in order to strengthen their differences, the HV-MSGD is combined with bilateral filtering (BF), which can enhance the target intensity while suppressing the background. Then the candidate target is extracted by adaptive local threshold segmentation and global threshold segmentation. The weak target is detected by the above single frame detection method. In order to further verify the effect on the single frame detection, the single frame detection algorithm mentioned above is combined with the improved unscented Kalman particle filter (UPF) to achieve the trajectory detection. The experimental results show that the proposed method is superior to other methods under weak signal-to-noise ratio (SNR), and can enhance the target while suppressing the background, and the enhancement effect is 6-30 times than other methods. In the experiment, the input SNRs are 2.78, 1.77, 1.79, 1.13 and 1.16, respectively. After image processing, the background suppression factors (BSFs) are 13.48, 21.33, 11.73, 20.63 and 121.92, and the signal-to-noise ratio gain (GSNRs) are 40.09, 71.37, 27.53, 12.65 and 131, respectively. The probability of detection (Pd) of this method is also better than other algorithms. The Pd for five sets of real sequence images is counted, when the false alarm rates (FAR) are 5×10^(-4), 1×10^(-3), 1×10^(-3), 1×10^(-5) and 7×10^(-6), the Pd are 94.4%, 92.2%, 91.3%, 95.6% and 96.7%.
keywords:dim target detection  multi-scale grayscale difference  distance and pixel difference  local threshold segmentation  global threshold segmentation
  HTML  查看/发表评论  下载PDF阅读器

版权所有:《红外与毫米波学报》编辑部