【作者】 史凌峰；

【导师】 郭宝龙；

【作者基本信息】 西安电子科技大学 ， 电路与系统， 2008， 博士

【摘要】 当前快速发展的无损探测技术已经被广泛应用在各个应用领域。作为探地雷达这种先进的无损探测应用技术显得尤为重要。这是因为探地雷达与电阻法、低频电磁感应法及地震法等常规的地下无损探测方法相比,它具有探测速度快、探测过程连续、分辨率高、操作方便灵活、探测费用低、探测范围广(包括金属和非金属)等优点,所以受到国内外广泛的关注。探地雷达中的信号处理技术是一项关键技术,现在成为国内外学者研究的热点问题。探地雷达信号处理包含众多的课题,其中埋地目标的成像、检测和识别技术是一项重要内容,这正是本论文所要完成的主要任务。　本文系统地研究了探地雷达的检测及其相关技术,在此基础上,针对探地雷达的重要环节,给出了一些新的技术,并实现了一个原理样机。本文的主要研究工作及创新点如下:　1.提出了一种分层均衡放大器的设计方法解决探地雷达成像对深度信号的衰减。该方法采用多测线平均得到回波的平均值,然后对原始测线组和平均回波进行分段,相当于对地下图像进行分层;根据其中最大的一次回波的峰峰值确定均衡放大的最大上限,将该最大值与平均回波各段的峰峰值的比值作为原始测线组的均衡放大矢量因子;使用均衡放大矢量因子对测线组进行矢量点积运算,从而得到均衡放大后的成像数据。该方法可以增强深度信号,利用计算机技术进行实时成像,就可以实现探地雷达深层探测和提高成像的分辨率。　2.提出了一种基于先验知识思想的非线性自适应插值算法(Nonlinearly　Adaptive　Interpolation,简称NLAI)以克服传统的固定插值算法的不足。该算法运用相邻节点的斜率来计算其二次导数,将其作为信号曲率大小判定依据,并根据插值数目的阈值判定插值数目,采取预测分段的方法获得信号的特征信息,从而选择特征级别对应的非线性插值函数,并重新优化分段,以达到最优插值的目的。通过该算法可以提高插值速度及插值效果,便于更好地观察探地雷达的实时波形。　3.提出了一种感兴趣目标区域插值算法(Interesting　Interpolation　Algorithm,简称为IIA)。IIA是通过构造被检测目标的相应模型,采用二次统计检测地下目标的水平位置,再用菱形模板预处理目标测线,用统计的方法自动检测出目标纵深位置,并对目标区域采用相关增强的方法,最后用双线性插值算法对感兴趣的目标区域进行放大而抑制背景,将结果成像出来达到感兴趣目标区域放大和细化。该算法克服采用人工检测识别方法中的判断目标位置比较模糊的局限性,且如果采取传统的图像增强技术,如锐化、直方图等技术,因为噪声及成像灰度级较少等原因而很难奏效。IIA可以实现埋地目标自动检测和感兴趣目标区域自动放大功能,消除人工判断目标区域的模糊性。　4.提出一种基于C扫描的基础上自动定位三维埋地目标的一种算法。该算法解决探地雷达对三维埋地目标自动定位问题。该算法采用并行变长滤波器组去除A扫描的直达波;其次对回波进行匹配小波奇异性检测获得目标是否存在的初步结果,检测方法采用基于统计意义下的多尺度小波域模极大值方法;再次对可疑位置采取时空域相关方法进一步确认以消除虚警概率;最后对得到的目标三维位置采取局部SAR聚焦的算法得到最终的目标位置。　5.提出一种采用高阶谱(Higher-Order　Spectra,HOS)作为目标识别特征依据的多特征融合方法识别三维埋地目标(简称MFR)。多特征包括埋地目标在测线方向的时间和空间域上呈献双曲线特征;三维埋地目标的深度特征,即衰减特性或者说幅度特性和埋地目标的空间相关性。高阶谱具有比功率谱更多的信息,主要是相位和频率信息,因此采用高阶谱作为目标识别的依据在理论上是可行的。从信号变换的角度看,高阶谱的分布特征间接反映了埋地目标对入射电磁波的脉冲响应。6.设计实现原理样机,通过实地实验测试所设计的无载波脉冲探地雷达可以实现对埋地目标的检测。该实验样机具有超宽带的天线、窄脉冲源和信号处理机的三个关键技术部分。更多还原

【Abstract】 The lossless detection technique is widely used in many fields of applications with its development. One of the untouched detection devices, namely ground penetrating radar (GPR), is an important to be used in practice. It possesses many properties than other routine, for instance resistance method, low frequency electromagnetic induction, seismic method, and so on. The properties are made up of fast detection, continuing detection, high resolution, convenient-flexible operation, low cost, wide detected range, and so forth. Therefore GPR is paid attention by people. It is a hot point to be studied that the signal processing of GPR belongs to a key technique in the system of GPR from internal to overseas. The signal processing of GPR includes many of problems, but imaging, detection and recognition about buried targets are important contents, and then they are discussed in the paper in detail.The paper studies GPR in the system. The detection and other associated techniques are done in GPR in detail. The paper presents a few of new techniques and realizes a theoretical sample of GPR based on the important parts of GPR. The studied work and innovative points are described in the paper as following.Firstly the paper proposes that a novel design method on layered uniform amplifier, which resolves the problem that deep signals of GPR are decayed. The method firstly takes many of measured lines that are averaged to obtain mean of echoes, and then both original-measured lines and averaged echoes are segmented, which is equal to make underground image layer. Secondly it determines uniform-magnified maximum based on maximum peak-peak value of echoes that is compared with the peak-peak value of averaged segments in echoes, and then the ratio is taken as uniform-magnified vector factor to original measured lines. Finally measured lines are operated with uniform-magnified vector factor by dot matrix to get layered-uniform imaging data. Therefore the method can enhance the signals in deep, and then computer technique is made full of imaging with real time to realize deep layer detection and improve resolution in GPR.Secondly the paper brings forward a new method of nonlinearly adaptive interpolation (NLAI) to overcome the limitation traditional fixed interpolation. NLAI firstly adopts slope of border upon nodes as its quadric derivative, which is acted as curvature determinant. So based the threshold of interpolated number, the characters of signal are obtained by pre-segment method in the algorithm. According to the nonlinear interpolation function, the algorithm optimizes segments to obtain optimal interpolation. Both the speed and effectiveness of the interpolation operation are improved by the NLAI. So the real-time echoes of GPR are observed better than without NLAI.Thirdly the paper puts forward an interesting area interpolation algorithm (IIA). IIA firstly structures a mode of the detected targets, and then adopts quadric statistic to detect the horizon position of the targets. Secondly it takes diamond templates to process measured lines that contain buried targets. Thirdly it detects the depth of buried targets by statistics. Fourthly IIA enhances the area of buried targets with the correlation. Finally it does double linear interpolation to magnify interesting area of targets, and then the results are imaged to obtain interesting area magnified and detailed. IIA overcomes the limitation of localizations by artificial detection and recognition. It is not effective that the traditional enhancing image techniques are adopted, for instance edge process, histogram enhancing image, and so on. The reason is that the gray level of imaging is few in GPR.Fourthly the paper presents a method that can locate buried three-dimension (3-D) targets automatically based on C-scan. The method is used to overcome the problem on 3-D targets automatically locating in GPR. It makes parallel filter groups eliminate direct arrival wave of A-scan with varied length. Then echoes are detected with matched wavelet to get its oddity, and then it simply determines buried targets whether or not. The detection method takes maximum mode based on statistics in multi-scale wavelet field. The suspicious positions are determined to decrease false alarm rate by space-time correlation method. Finally the positions of buried targets are obtained by local focus of synthetic aperture radar (SAR) on 3-D position.Fifthly the paper presents a method that recognizes 3-D buried targets based on both high-order spectra (HOS) distributing features and multi-feature recognition (MFR) method. Multi-features of GPR include the hyperbola in space-time along measured lines, the depth feature of 3-D buried targets, namely attenuation waves or varied scope of echoes, and correlation of buried targets in space. HOS possesses more properties than power spectrum, for example phase and frequency information. HOS is acted as the rule of recognized targets so that it is feasible in theory. Thus the response of incidence electromagnetic wave is mapped indirectly to the distributing of HOS on buried targets.Sixthly the theoretical sample of pulse ground penetrating radar without carrier is completed in the paper. The device is proved to detect buried targets by practical tests. The designed sample possesses key techniques that are ultra-wide antenna, narrow pulse resource and signal processor.更多还原