【作者】 王海霞；

【导师】 李琦；

【作者基本信息】 哈尔滨工业大学 ， 物理电子学， 2006， 硕士

【摘要】 大量的军用和民用需要识别位于未知场景中静止或运动的目标,可以利用多种传感器来进行目标识别,如:3-D激光雷达、视频传感器、前视被动红外系统等等。前视被动红外传感器较激光成像雷达作用距离远,且因是被动系统,所以工作时不会暴露自身的位置,但其抗干扰能力差、图像不稳定,而且探测虚警概率高。激光成像雷达能成距离像和强度像,可以很好地完成目标辨认任务,但其缺点是作用距离较近。本文主要针对3-D激光雷达主动成像的目标检测和目标方位角估计进行理论研究和计算机仿真,着重阐述了3-D激光雷达主动成像的目标检测理论和基于李群的希尔伯特-施密特界(HSB)理论。首先对3-D脉冲相干成像激光雷达系统作了简单的介绍,并且给出了针对该激光雷达系统进行的理论分析,给出了被动红外成像单像素统计模型。然后给出了3-D脉冲相干激光雷达距离像的仿真思路,利用MATLAB,计算了对目标的大小、载噪比、距离分辨率分别取不同的值时的探测概率的变化,并给出了各个参数和探测概率的关系曲线。其次对希尔伯特-施密特界进行了介绍,主要集中在自动识别系统中的目标可变形模板的表达方式和群作用的表示方法上。在这种框架中,利用李群来表示旋转空间,然后基于贝叶斯估计结构,计算了利用李群的表示的最小均方误差估计性能界限。对文献中已给的目标方位角的最小均方误差表达式进行了推导,并且给出了方位角的最小均方误差和噪声偏差之间的关系。最后,计算了简单情况下关于目标方位角的希尔伯特-施密特界,并给出了方差和HSB之间的关系曲线。更多还原

【Abstract】 Avariety of civilian and military applications require recognizing stationary or moving objects, situated in unknown surroundings, using standard sensors such as three-dimensional laser radar、video sensor and forward-looking infrared sensors. The range value of the Forward-looking infrared sensors is father than laser radar. But a Forward-looking infrared sensor is a passive system, so its position can easily be discovered, the image is very instability and the false alarm probability is very high. Laser radar can present range image and intensity image, so it can accomplish the detection mission, but it also have flaw, the function of the range is very near.This paper aim at the theory and computer emulator of detecting object that use three-dimensional laser radar and the estimation of object’s pose. And emphasize the theory of three-dimensional laser radar and Hilbert-Schmidt Bounds (HSB) for estimators on matrix Lie groups.First this paper introduces a three-dimensional laser radar and analyses the laser radar system, then introduces the forward-looking infrared (FLIR) system and its single pixel statistics. Then give the emulator thought of the three-dimensional laser radar range image. Using MATLAB calculate the relation between the objects cross section、carrier-to noise ratio、the range resolution and probability of detection, then paint relation curve of the parameters and the probability of detection.Secondly, this paper introduces the Hilbert-Schmidt Bounds (HSB) and mostly focuses on the express fashion of formable objects template and express fashion of matrix Lie group of automatic target recognition (ATR) system. Upon this frame, use matrix Lie group to express the rotation space, then base on Bayesian estimation framework, calculate the minimum mean squared error bounds which use the matrix Lie groups. Extend the expressions of the minimum mean squared error bounds of the references, and give the relations of the minimum mean squared error bounds of pose angle and noise standard deviation. In the end, calculate minimum mean squared error bounds which use the matrix Lie groups to express the object’s orientation in the simple instance, and paint更多还原