MBM01和δ-GLMB

1. 在预测步骤中，δ-GLMB 需要通过 K 最短路径算法截断预测密度，而 MBM 则很简单。
2. 在更新步骤中，由于一个MBM全局假设可以有效地表示多个δ-GLMB全局假设，因此PMBM具有较少的全局假设，这将导致更有效的数据关联。具体来说，一个包含 n 个伯努利分量的多伯努利密度可以表示为具有 2ⁿ 个全局假设的 MBM01 或 δ-GLMB。

3. PMBM的泊松出生假设有效地代表了潜在的目标。然而，由于 δ-GLMB 密度中的每个全局假设都具有确定性基数，因此 δ-GLMB 参数化需要无限数量的全局假设来表示泊松部分。

1. In the prediction step, δ-GLMB needs to truncate the predicted density by a K-shortest path algorithm, whereas MBM is straightforward.
2. In the update step, since one MBM global hypothesis can efficiently represent multiple δ-GLMB global hypotheses, the PMBM has fewer global hypotheses, which will result in more efficient data association. Specifically, one multi-Bernoulli density containing n Bernoulli components can be represented as MBM01 or δ-GLMB with 2ⁿ global hypotheses.

3. The Poisson birth assumptions of the PMBM effectively represent potential targets. However, since each global hypothesis in the δ-GLMB density has a deterministic cardinality, δ-GLMB parameterization would need an infinite number of global hypotheses to represent the Poisson part.

 

 感谢 https://www.cnblogs.com/sunny99/ sumoier对本文的帮助！