Spatio-temporal credible evidence fusion for UAV type recognition in distributed urban low-altitude sensing networks

In continuous recognition of urban low-altitude unmanned aerial vehicles (UAVs), the widespread deployment of low-cost sensors and autonomous detection technologies has led to multi-view, high-dimensional data exhibiting strong temporal correlations. However, large-scale evidence that exhibits spatio-temporal inconsistencies caused by environmental clutter, building occlusion, and measurement errors leads to high computational complexity, underutilization of informative data, and degraded system stability. Decision-level fusion methods relying on pairwise comparison metrics impose a high computational burden and underutilize the contributions of high-quality evidence. The lack of an effective measure for temporal evidence consistency fluctuations restricts system performance in achieving efficient, accurate, and stable recognition. To address these challenges, this paper proposes a spatio-temporal credible evidence fusion method, STCEF, for distributed UAV type recognition in urban low-altitude within a collaborative spatio-temporal fusion framework. Spatially, a conditional credibility-based fusion method under a fusion-center-based strategy is designed to reduce computational complexity while improving the utilization of high-quality evidence. Temporally, sliding-window Rényi entropy quantifies uncertainty fluctuations to dynamically adjust evidence fusion weights, thereby suppressing transient interference caused by variations in class and belief assignments. Simulation results demonstrate that the proposed method enhances computational efficiency while achieving more accurate and stable type recognition, improving accuracy by 2.1% and mean belief by 10% compared with conventional methods.