Study On Building Layout Reconstruction Techniques Based On Graph Theory From Through-the-wall Radar

Electromagnetic(EM) waves especially those with low frequency have the capabilities of penetrating through nonmetallic building materials using, it is provided with a unique opportunity to image, detect and locate target behind obstacle such as walls.Through-the-wall radar(TWR) which is based on this principle has been widely applied in disaster rescue action, counter terrorism and riot as well as urban street fighting in recent years. Building interior layout reconstruction radar also belongs to TWR. However,it is different from the traditional TWR which requires a close distance or directly face to the wall. Building interior layout reconstruction radar is able to observe a whole building from stand-off distance, wide view angle and multiple directions. TWR echo data or image are analyzed in order to reconstruct the building interior layout. Research work on reconstruction algorithm has only just begun worldwide. This dissertation focuses on this important issue. A building layout graph(BLG) model has been established whose node set and edge set is made up by wall-wall-floor trihedrals and walls, respectively. We propose a graph-theoretic based method to reconstruct the building interior layout. This dissertation includes the following main achievements and contributions.Firstly, how to acquire the location attributes of the nodes is researched. We have proposed an image-domain estimation method for calibration parameters of node location.We have translate the problem of estimating the calibration parameters into optimizing the parameters of a filter which is designed in image domain. Thus, the delay effect of wall on EM waves is compensated, providing more accurate location coordinates of principle scatterers inside the building. The image-domain based method not only supplies high estimation accuracy but also improves computation efficiency. After that we have made detailed theoretical analysis on polarimetric characteristics of principle scatterers inside a four-wall formed rectangular building. Electromagnetic modeling experiments and anechoic chamber experiments have demonstrated our conclusions. These polarimetric characteristics to distinguish principle scatterers such as dihedral and trihedral. Then we use classical constant false alarm rate method and morphological filter to extract region of interest of the principle scatterers in building radar image. Thereby, the location attributes of the BLG nodes are obtained.Secondly, we have studied the method for acquiring the angle attributes of the nodes. The pose angle of the trihedral formed by the wall-wall-floor structure inside the building is very important in mapping its interior layout. In this paper, an image-domain based method is proposed to estimate the pose angle of the trihedral using a feature called amplitude ratio(AR). The estimated pose angle of a trihedral is determined according to AR. After the imaging geometry of the radar with a multiple-input multiple-output(MIMO) array and the definition of AR in the echo-domain are described, a parametric backscattering model based on geometrical theory of diffraction is applied to analyze AR in the echo-domain when a trihedral is in different pose angles. The GTD model can be treated as the analytical matching templates library. Considering the echo-domain method can only estimate the pose angle of one trihedral, a virtual aperture imaging model which describes the imaging procedure using MIMO array is developed in this dissertation to deal with the case of multiple trihedrals inside a building. Based on the imaging model, the AR of each trihedral can be calculated in the image-domain instead of the echo-domain,overcoming the deficiency of the echo-domain method.Thirdly, after we have obtained the location attributes and the angle attributes of the nodes, a method based on minimum spanning tree(MST) which comes from graph theory is proposed in order to reconstruct the building interior layout from through-the-wall radar image, solving the automatic computation problem in iterative reconstruction procedure.Definition of the weight between any two nodes of BLG is significant in applying the MST-base method. This dissertation makes full use of the location and angle attributes to construct reasonable edge weight. Moreover, we have demonstrated that the MST of BLG is completely equivalent to the correct building interior layout. By this way, we can translate the building interior layout reconstruction process into searching the MST of the complete weighted undirected graph BLG. The MST searching method in graph theory is employed to reconstruct the simulated building interior layout.Afterwards, we have applied the traditional MST method to process electromagnetic modelling data which is simulated in omni-directional observation and ortho-directional observation, respectively. The results have demonstrated the suitability and efficacy of the iterative MST-based reconstruction method. Considering there will exist interference targets inside the building in practical measurement, the corner-point and wall-edge information in the radar image are utilized to improve the traditional MST method. We have established a MST model with restricted conditions on edge and proposed construction method for MST in this case. By using this method, the interference targets can be removed to obtain the correct building interior layout. The improved MST-based method can also be applied to the reconstruction case when there exist no interference targets.Finally, a summary of the dissertation is made, while several open problems for the building interior layout reconstruction are proposed.