Automated regularity assessment of UAV swarm spatial structure is essential for safe swarm control, yet existing distance-based methods are vulnerable to kinematically coupled false positives and unstable decisions near regime boundaries. To address these challenges, this paper first introduces a Spatiotemporal Coupled Dynamic Association Matrix (SCDAM), which incorporates velocity-direction consistency into pairwise distance relationships to distinguish geometrically similar but dynamically different swarm states, thereby improving the reliability of spatial representation. Building upon this dynamic representation, an FFT-based spectral analysis framework is further developed to extract multidimensional regularity descriptors from autocorrelation spectra, achieving robust feature characterization using FFT-based computation. To improve sequence-level decisions, a dual-threshold hysteresis mechanism with an emergency bypass is proposed to suppress unnecessary state switching while preserving rapid responses to abrupt behavioral changes. Experiments on the UCI Swarm Behaviour dataset and physically parameterized offline trajectory/interpolation simulations show higher static accuracy, improved temporal stability, and robustness to the tested coordinate transformations and measurement disturbances. These results support the feasibility of the proposed assessment pipeline under the evaluated conditions; real-flight, onboard sensing, communication, and actuator validation remain future work.



