GPR TARGET RECOGNITION FRAMEWORK BASED ON LOW-RANK SPARSE DECOMPOSITION AND HYPERBOLIC SCANNING

As a non-destructive detection technology, ground penetrating radar (GPR) has been widely used in the target detection of roads and walls. Strong clutter interference is an important obstacle to affect the accuracy of target recognition, and how to accurately extract the target from the GPR images is a challenging task. This paper presents a target identification framework for GPR. In this method, the low-rank sparse decomposition (LRSD) method is used to extract the target part from the radar image. This method uses the alternating direction multiplier method (ADMM) for iteration. The dictionary learning method uses the dictionary to further process the image with the downward opening. Finally, the hyperbolic point linearization method is used to identify the hyperbola. The proposed method achieves good results in the measured data.