Knowledge-aided Object-oriented Three-dimensional Microwave Imaging

Recent years has witnessed a growing interest in the research for featured-enhanced high-resolution microwave imaging field. The existing methods for radar image process are classified into image processing and target classification. Though these methods can effectively improve the imaging quality, the fact we have to meet is that lots of useful image information would be discarded in the procedure. For solving this problem, a closed-loop framework involved in object-oriented microwave imaging method using prior knowledge is introduced in this paper, which results in decreasing the computation cost and enhancing reconstructions resolution. The proposed methodology has several key advantages: e.g., this proposed method converts the conventional radar imaging problem into physic-driven image processing problem via a proximal operator, which admits us to apply prior knowledge into the algorithm procedure helping greatly improve the imaging quality and accelerate the imaging speed. Furthermore, a positive feedback is exploited between target classification and image processing stage. More prior knowledge on object is brought from the feedback applied into the imaging procedure. In this way, the reconstruction quality is significantly improved. What is more, the positive feedback could also significantly enhance the robustness of reconstruction and the probability in target classification. Selected simulation results are provided to demonstrate the state-of-art performance of the proposed methodology.