基于时空域滤波的红外小目标检测

发布时间：2018-07-27 10:30

【摘要】：为了提高红外小目标探测能力，本文以地球静止轨道凝视探测器对地探测为应用背景，提出了两种对小目标进行探测的算法：三维匹配滤波算法和基于投影算法的目标检测方法。针对动平台固有抖动性，凝视探测器拍摄到的序列图像中目标关联性低这一问题，提出一种新的抖动补偿算法。本文的主要研究工作如下：（1）分析了红外小目标图像特性，给出了算法仿真输入的理论依据；（2）根据红外云图的背景特性，采用最小二乘法原理构建了基于灰度的背景模型，以用来对序列图像的抖动量进行估计，并对序列图像配准；（3）利用最大信噪比原则和柯西—施瓦茨不等式推导了三维匹配滤波器的一般数学公式。对于序列图像，给出了三维匹配滤波算法的一般流程并且进行了算法仿真，有效地验证了三维匹配滤波算法对于低信噪比小目标的探测具有非常优越的性能，但是三维匹配滤波算法对于速度失配的运动目标将表现的很差，其要求更加密集的速度方向滤波器，对现有计算机存储和计算能力提出了较高要求；（4）针对经投影算法计算得到的二维图像，设计了一种引入分层投票理论的非关联点移除算法，有效地对孤立点和弱相关点去除，保留了目标轨迹；通过引入hash原理设计计算了对多目标轨迹进行分类的参数，，有效地将多目标轨迹进行分类，并且给出了仿真；（5）利用本文研究成果，设计开发了基于GUI的红外小目标检测软件系统，可以实现抖动序列生成，待配准帧抖动量计算，三维匹配滤波算法对小目标检测和基于投影算法的目标检测。本文所研究的基于时空域滤波的红外小目标检测可以为在轨目标检测提供理论基础。
[Abstract]:In order to improve the detection ability of small infrared target, the geostationary gaze probe is used as the application background in this paper. Two algorithms are proposed to detect small targets: 3D matched filter and projection based target detection. In view of the inherent jitter of the moving platform and the low correlation of the target in the sequence images taken by the staring detector, a new jitter compensation algorithm is proposed. The main research work of this paper is as follows: (1) the characteristics of infrared small target image are analyzed, and the theoretical basis of algorithm simulation input is given. (2) according to the background characteristics of infrared cloud image, The background model based on gray level is constructed by using the least square method to estimate the jitter of the sequence image and to register the sequence image. (3) based on the principle of maximum signal-to-noise ratio (SNR) and Cauchy Schwartz inequality, the general mathematical formula of 3D matched filter is derived. For the sequence images, the general flow of 3D matched filtering algorithm is given and the algorithm simulation is carried out, which effectively verifies that the 3D matched filtering algorithm has very superior performance for detecting low SNR small targets. But the 3D matched filtering algorithm will be very poor for the velocity mismatched moving target, which requires more intensive velocity direction filter, and puts forward a higher demand for the existing computer storage and computing power. (4) for the two-dimensional images calculated by the projection algorithm, a non-correlation point removal algorithm based on the hierarchical voting theory is designed, which can effectively remove the isolated points and weak correlation points and keep the target track. By introducing the principle of hash, the parameters of multi-target trajectory classification are designed and calculated, the multi-target trajectory is effectively classified, and the simulation is given. (5) using the research results of this paper, The infrared small target detection software system based on GUI is designed and developed, which can generate the jitter sequence, calculate the jitter amount of the frame to be registered, detect the small target with 3D matched filtering algorithm and detect the target based on projection algorithm. In this paper, the small infrared target detection based on spatio-temporal filtering can provide a theoretical basis for in-orbit target detection.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751

【参考文献】