Multi-Domain and Multi-Dimensional Feature Modeling and Analysis of Low， Slow， and Small Targets via Ubiquitous Radar Under Sea and Air Background

‍ ‍The weak echo and indistinct features of low， slow， and small （LSS） targets， such as birds and drones， pose high requirements for radar detection and recognition. Feature modeling and analysis are the foundation， while obtaining multi-domain and multi-dimensional target features for radar is the prerequisite. The digital array ubiquitous radar achieves long-time integration of targets through the “wide beam transmission and narrow beam reception” working mode， achieving a higher integration gain and Doppler resolution. It can obtain multi-dimensional features of targets in multiple domains， laying the foundation for the fine processing of LSS targets and the integration of detection and recognition. This study focuses on LSS targets such as birds， rotors， fixed-wing aircraft， and helicopters in sea and air environments. Using multi-dimensional data such as range， azimuth， inter-frame and range， pulse， inter-frame obtained from a digital array ubiquitous radar system， time-domain echo features of the LSS targets （single frame and dynamic pulse echoes）， transform domain Doppler features （Doppler waterfall plot and micro-Doppler spectrum）， and long-time maneuvering features （acceleration sequence， and trajectory） 7 categories and 19 types of multi-domain and multi-dimensional features are extracted， which can fully reflect the amplitude fluctuations， energy changes， motion， maneuvering， micro-motion， and other characteristics of the target between single and multiple frames of data， thereby achieving fine characterization and analysis of LSS targets. Finally， a dataset of LSS target features was collected and constructed using the ubiquitous digital array radar. The characteristics of typical targets were validated and quantitatively and qualitatively analyzed， and the feature differences of different targets were summarized. Regarding signal characteristics， single-frame pulse echo plots of the rotor wing drone， helicopter， and fixed-wing aircraft are periodic； the single-frame pulse echo of the flying bird fluctuates irregularly and is not periodic. The time-domain information entropy of the echoes of the flying bird is the largest， followed by that of the helicopter and the fixed-wing airplane， and that of the rotor-wing drone is the smallest. The radar cross section （RCS） is the largest for helicopters， followed by flying bird flocks， and the smallest for fixed-wing aircraft and rotor-wing drones. The micromotion characteristics of helicopters， rotor-wing drones， and fixed-wing aircraft are obvious and have periodic variations； those of flying bird flocks have some micromotion characteristics but do not have periodic variations. As for the motion characteristics， the motion of the aircraft is relatively smooth， and the changes in speed and acceleration are also relatively smooth， while the maneuvering of the flying birds is stronger and more irregular. Helicopters and fixed-wing aircraft basically do a straight line or simple curve motion； flying birds do a nearly straight line or curve motion with small curvature， while rotor-wing drones are manipulated by human beings， and the corresponding trajectories are more complicated. The validation results showed significant differences in the multi-dimensional features for the four types of LSS targets， and the obtained features and differences would provide important support for subsequent classification and recognition of LSS targets.