Research On Key Techniques For Through Wall Radar Imaging

Through wall radar imaging (TWRI) is a newly developed technique that can form an image of region of interest which is behind an obstacle. Owing to its capabilities of detection, imaging, positioning, tracking and recognition of the target that is located behind the wall, TWRI has been involved in a wide range of applications, such as street battle, anti-terrorist, disaster relief and hostage rescue. As a new technique applied in a new environment, TWRI is confronted with some difficulties and challenges. Based on the study of basic principles and application characteristics of TWRI, this paper carries out the research on several key problems for the purpose of producing a high-quality through wall radar image.At first, the technique for suppression of wall clutter is studied. Wall clutter usually has a large amplitude and a long time duration, tending to cover up target echoes, thus it has to be suppressed. In the third chapter, this thesis proposes two methods, which are performed in echo domain and image domain, respectively, to mitigate the wall clutter. By taking advantage of that wall clutter varies slowly as the antenna position changes, the echo domain approach combines the SaS and SVD techniques to suppress wall clutter. Based on the fact that the components of a target pixel are consistent while the components of a clutter pixel are not, the image domain approach introduces the Coefficient of Variation to weight the pixel for enhancing the target pixel while suppressing the clutter pixel.Then the techniques for elimination of multipath ghost are investigated. In indoor environment, since the imaging scene is enclosed by four walls, the radar signal will not only propagate along the direct path between the radar and target, but also propagate along other paths which include reflections at interior walls. The signal propagating along these paths are termed as multipath echoes, which will produce multipath ghost in the resulting image, causing the increase of false positives. To solve this problem, a multipath model is established firstly in the fourth chapter. Then based on this model, a method that eliminate the ghost by cancelling multipath echoes is proposed. In this method, time delays and amplitudes of multipath echoes are estimated, then multipath echoes can be subtracted from the received radar returns. Without multipath echoes, the ghost will not appear in the newly formed image any longer. In addition to this method, a ghost suppression approach using compressive sensing (CS) technique is presented. In this approach, each multipath is viewed as an observation channel and is expressed by an observation matrix. Then these matrices are assembled into one observation framework, based on which the CS algorithm can be used to generate the imaging result. Since each multipath echo has been seen as the echo of the target under another observation, the multipath echoes will be projected to the position of the real target, avoiding the appearance of the ghosts.Next, this thesis deals with the problem of the absence of wall parameters in TWRI. In practice, wall parameters are usually unknown and can not be measured in advance. If the incorrect wall parameters are used, the resulting image will be blurred and smeared. To overcome this difficulty, two solutions are provided in the fifth chapter. The first one is to estimate the wall parameters before imaging procedure and this thesis proposes a SVR-based method to achieve the estimation result. By taking a large amount of wall echoes with known wall parameters as training samples, the method employs SVR to train regression functions. When a new echo from a wall without known parameters is received, the echo is input into the regression functions, from which the outputs are used to acquire the estimation result through minimizing a predefined cost function. The second is to use autofocusing technique to produce the image without wall parameter estimation. Since the normal autofocusing technique is computationally expensive, a simple and fast autofocusing method is proposed by simplifying the image quality criterion and reducing optimization variables. This method can obtain the imaging result that is similar with that produced under known wall parameters while improving the computational efficiency significantly.At last, the fusion imaging and image fusion methods for TWRI are discussed. In TWRI, the imaging scene usually can be observed from multiple viewing angles or by multiple observation modes, and the multiple observation results can be fused to provide a higher quality and more informative image. For the situation that the positions and amplitudes of scattering points can be viewed as invariant during two observations, a fusion imaging method based on CS technique is proposed in the sixth chapter. Since the positions and amplitudes of scattering points are viewed as invariant, the data from two observations can be put into one observation model, from which a fine resolution image without high sidelobes can be obtained using CS algorithm. For the situation that the positions and amplitudes of scattering points can not be viewed as invariant during two observations, a double-layer fuzzy fusion method is proposed to fuse the images provided by each observation. This method exploits the information of the target within each single image in the first layer that is called as intra-image fusion, and then based on the fusion result of the first layer, inter-image fusion is performed to fuse the information between two images in the second layer. The simulation and experimental results both prove that this method can enhance the target and suppress the clutter, producing a better through wall imaging result.