Causal neural set filtering for online multi-target tracking
Causal neural set filtering for online multi-target tracking
用于在线多目标跟踪的因果神经集滤波
Abstract: Transformer-based multi-target tracking (MTT) jointly learns data association and state estimation, but MT3/Track-MT3-style trackers repeatedly re-encode measurement windows, incurring redundant computation.
摘要： 基于 Transformer 的多目标跟踪（MTT）联合学习数据关联和状态估计，但 MT3/Track-MT3 风格的跟踪器会重复编码测量窗口，从而导致冗余计算。
We propose Causal Neural Set Filtering (CNSF), a neural set filter that encodes only current measurements while carrying past evidence in a structured recursive track state.
我们提出了因果神经集滤波（CNSF），这是一种神经集滤波器，它仅对当前测量值进行编码，同时将过去的证据保留在结构化的递归跟踪状态中。
CNSF combines exclusive Sinkhorn association, association-conditioned Kalman-shaped updates with moment matching, and recurrent Bernoulli lifecycle modeling with measurement-driven birth.
CNSF 结合了排他性 Sinkhorn 关联、基于关联条件的卡尔曼式更新（带矩匹配），以及具有测量驱动生成机制的循环伯努利生命周期建模。
These mechanisms impose soft one-to-one constraints, propagate association-induced state uncertainty, and support existence estimation under missed detections and birth—death transitions.
这些机制施加了软性的一对一约束，传播了由关联引起的状态不确定性，并支持在漏检和生灭转换情况下的存在性估计。
On a held-out three-regime simulated test set, CNSF reduces mean GOSPA and T-GOSPA relative to Track-MT3 by 19.3% and 30.4%, with 55.9% fewer parameters and a $3.76\times$ speedup in single-thread CPU inference.
在保留的三状态模拟测试集上，与 Track-MT3 相比，CNSF 将平均 GOSPA 和 T-GOSPA 分别降低了 19.3% 和 30.4%，参数量减少了 55.9%，且单线程 CPU 推理速度提升了 3.76 倍。