基于图理论的穿墙雷达建筑物布局重构技术研究

发布时间：2018-06-14 19:38

本文选题：穿墙雷达 + 图理论　；参考：《国防科学技术大学》2014年博士论文

【摘要】：电磁波特别是低频电磁波对非金属建筑材料具有穿透特性,它可以对隐蔽于墙体后面的目标进行成像、探测和定位,基于此原理的穿墙雷达(Through-theWall Radar,TWR)近年来在灾难救援、反恐防暴以及城市巷战等领域得到了广泛应用。建筑物布局重构雷达属于TWR的一种,但其有别于需贴墙或离墙很近工作的传统穿墙雷达,它可以远距离大视角多方位对整栋建筑物进行观测成像并通过对穿墙雷达回波和图像的处理达到重构建筑物布局的目的。国内外对于建筑物布局重构算法的研究才刚刚起步,本文围绕此问题构建了以墙-墙-地板形成的介质三面角为节点、以墙为边的建筑物布局图(Building Layout Graph,BLG)模型,提出了一种基于图理论的建筑物布局重构方法,主要研究工作和创新点如下:首先,研究了节点位置属性获取方法。提出了一种基于图像域的节点位置校正参数估计方法,通过在图像域设计滤波器,将节点位置校正参数的估计问题转化为滤波器参数的最优化问题,进而补偿掉因墙体存在引起的电磁波传播延时效应,保证获得尽量精确的建筑物内部典型散射体位置坐标,该方法在保证了较高估计精度的同时也提高了运算效率。校正参数估计完成后对四面墙构成的矩形建筑物内部的典型散射体极化特性进行了详细的理论分析、电磁仿真验证和暗室试验验证,利用极化特性对二面角、三面角等典型散射体进行分类后,再结合恒虚警(Constant False Alarm Rate)方法和形态学滤波提取典型散射体的感兴趣区域(Region of Interest),进而得到BLG中节点的位置属性。其次,研究了节点角度属性获取方法。由墙-墙-地板结构形成的三面角姿态对于确定正确的建筑物布局非常重要,本文基于三面角节点的幅度比(Amplitude Ratio,AR)特征提出了一种在图像域对其姿态角进行估计的方法,姿态角的估计值最终由计算出的AR来确定。在分析了多输入多输出(Multiple-Input MultipleOutput,MIMO)阵列的虚拟孔径雷达成像几何并给出AR的定义之后,本文结合基于几何绕射理论(Geometry Theory of Diffraction)的参数化后向散射模型先在回波域分析AR,以此作为解析的匹配模板库。考虑到实际建筑物内部往往存在多个三面角,而在回波域只能估计单个三面角姿态的情况,本文利用一种描述MIMO阵列成像过程的虚拟孔径成像模型在图像域对建筑物内部所有的三面角节点角度属性进行估计,有效克服了回波域方法的不足。在获取了节点的位置属性和角度属性以后,本文利用图理论方法研究了基于最小生成树(Minimum Spanning Tree,MST)的建筑物布局重构方法,解决迭代重构过程中的自动计算问题。BLG的边权值定义是运用最小生成树方法的关键,本文充分利用三面角节点的位置属性和角度属性构造了合理的边权值,并证明了权值定义中的参数在一定范围内取值时,BLG中的最小生成树与建筑物布局是完全等价的,从而将建筑物布局重构过程转化成赋权完全图BLG中最小生成树的搜索过程,再利用图理论中的最小生成树搜索算法对仿真的建筑物布局进行了重构。随后,本文利用最小生成树方法处理了全视角观测和正交视角观测两种情况下典型建筑物的高频近似电磁仿真数据,验证了基于图理论的迭代重构方法的有效性。对实际测量中建筑物内部容易出现干扰目标的情况,本文充分利用雷达图像当中的角点信息和墙边信息,对传统的最小生成树方法进行了改进,构建了一种带边约束条件的最小生成树模型,并提出了此约束条件下的最小生成树构造方法,以此为基础剔除掉干扰目标获得正确的建筑物布局,改进的最小生成树方法同样适用于不存在干扰目标条件下的建筑物布局重构。最后,对全文进行总结,并指出建筑物布局重构问题中需进一步研究的方向。
[Abstract]:Electromagnetic waves, especially low frequency electromagnetic waves, have penetrability to non-metallic building materials. It can imaging, detecting and locating targets hidden behind the wall. Through-theWall Radar (TWR) based on this principle has been widely used in disaster relief, anti terrorist riot and urban street warfare in recent years. The layout reconfiguration radar is a kind of TWR, but it is different from the traditional through wall radar that needs to stick to the wall or close to the wall. It can observe and imaging the whole building from a long distance from a large angle to the whole building and reconstruct the layout of the building through the processing of the echo and image of the wall. The research of the algorithm is just starting. In this paper, a new method of building layout reconstruction based on Building Layout Graph (BLG) is proposed. The main research work and innovation points are as follows: first, the nodes are studied. A method of location attribute acquisition is proposed. A method of parameter estimation is proposed based on the image domain. By designing a filter in the image domain, the estimation of the position correction parameter is converted to the optimization problem of the filter parameters, and then the delay effect of the electromagnetic wave propagation caused by the existence of the wall is compensated. The position coordinates of the typical scattering body inside the building are accurate. This method ensures high estimation accuracy and improves the operation efficiency. After the correction parameter estimation is completed, the polarization characteristics of the typical scatterers in the rectangular building made up of the four wall walls are analyzed in detail, the electromagnetic simulation verification and the dark room test are verified. After classification of typical scatterers such as dihedral angle and three plane angle with polarization characteristics, the region of interest (Region of Interest) of typical scatterers (Region of Interest) is extracted with Constant False Alarm Rate method and morphological filtering, and then the location attributes of nodes in BLG are obtained. Secondly, the acquisition method of node angle attributes is studied. The three plane angle posture formed by the floor structure is very important to determine the correct layout of the building. Based on the amplitude ratio (Amplitude Ratio, AR) of the three surface angle nodes, this paper proposes a method to estimate the attitude angle in the image domain. The estimation value of the attitude angle is determined by the calculated AR. After the Multiple-Input MultipleOutput, MIMO) array's virtual aperture radar imaging geometry and the definition of AR, this paper analyzes the AR in the echo domain based on the parameterized backward scattering model based on the geometric diffraction theory (Geometry Theory of Diffraction), which is used as the matching template library for the analysis. In this paper, a virtual aperture imaging model describing the MIMO array imaging process is used to estimate the angle attributes of all three angle nodes in the building, which can effectively overcome the shortage of the back wave domain method. The position of the node is obtained. After the attribute and the angle attribute, this paper uses the graph theory method to study the building layout reconstruction method based on Minimum Spanning Tree (MST). To solve the automatic calculation problem in the iterative reconstruction process, the edge weight definition of.BLG is the key to use the minimum spanning tree method. This paper makes full use of the location attribute of the three side angle nodes. And the angle attribute constructs reasonable edge weights, and proves that the minimum spanning tree in BLG is completely equivalent to the building layout when parameters in the definition of weight value are taken in a certain range. Thus, the reconstruction process of the building layout is transformed into the search process of the smallest spanning tree in the empowerment complete graph BLG, and then the minimum generation in the graph theory is used. The tree search algorithm reconstructs the simulation building layout. Then, this paper uses the minimum spanning tree method to deal with the high frequency approximate electromagnetic simulation data of the typical buildings under the full angle observation and the orthogonal visual angle observation in two cases, and verifies the validity of the iterative reconstruction method based on the graph theory. It is easy to interfere with the target. This paper makes full use of the corner information and wall information in the radar image, improves the traditional method of the minimum spanning tree, constructs a minimum spanning tree model with edge constraint conditions, and puts forward the minimum spanning tree construction method under this constraint, which is based on the elimination of dry tree. In order to get the correct layout of the building, the improved minimum spanning tree method is also applicable to the reconstruction of the building layout without interference target. Finally, the full text is summarized, and the direction of further research in the reconstruction of the building layout is pointed out.
【学位授予单位】：国防科学技术大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】