Joint Detection and Recognition of Satellites’ Orientated Components with Compact Polarimetric ISAR

Compared with optical imaging system, polarimetric inverse synthetic aperture radar (ISAR) can work all day and all-weather, which plays an important role in space surveillance. Accurate identification of the structural components of man-made targets can yield valuable information for distinguishing target behaviors. In this vein, an anchor-free orientated component detector is established. The orientated bounding box of target component is modeled by a 2-D Gaussian distribution. Consequently, the angle, center point position and size parameters can be coupled for joint optimization. Additionally, since there is lack of public dataset available, compact polarimetric ISAR orientated component annotation datasets are constructed for five satellites. The experimental results demonstrate the better performance for the proposed method with the mean average precision (mAP) index of 0.618.