# (ICIP2017)SIMPLE ONLINE AND REALTIME TRACKING WITH A DEEP ASSOCIATION METRIC

## ICIP2017, MOTS

Posted by pshow on June 16, 2020

## （ICIP2017）SIMPLE ONLINE AND REALTIME TRACKING WITH A DEEP ASSOCIATION METRIC

### 主要贡献

Simple Online and Realtime Tracking (SORT)

• 集成外表信息来提升SORT的性能

integrate appearance information to improve the performance of SORT

• 可以追踪被产时间遮挡的目标，减少 ID switch

we are able to track objects through longer periods of occlusions, effectively reducing the number of identity switches

• 在大的人物re id 数据集上训练相关矩阵

we place much of the computational complexity into an offline pre-training stage where we learn a deep association metric on a largescale person re-identification dataset.

• 使用 measurement-to-track associations 在视觉空间内进行最邻近查询

we establish measurement-to-track associations using nearest neighbor queries in visual appearance space.

• 减少了45%的 ID switch

### 简介

object trajectories are usually found in a global optimization problem that processes entire video batches at once.

due to batch processing, these methods are not applicable in online scenarios where a target identity must be available at each time step.

Simple online and realtime tracking (SORT)

SORT会引入大量的ID switch， 因为关联矩阵只有在状态评估置信度高的情况下才准确。

SORT 不能很好地处理遮挡问题。

### 利用深度关联矩阵来SORT（SORT WITH DEEP ASSOCIATION METRIC）

#### 追踪句柄与状态估计

The track handling and Kalman filtering framework is mostly identical to the original formulation in [12].

We use a standard Kalman filter with constant velocity motion and linear observation model,

$(u,v,\gamma,h, \dot{u}, \dot{v}, \dot{\gamma},\dot{h})$ 8维向量来表示跟踪目标的状态，bbox中心坐标，宽高比，高度，及其对应的速度。

$(u,v,\gamma,h)$ 为目标状态的直接观测值。

During this time, we expect a successful measurement association at each time step. Tracks that are not successfully associated to a measurement within their first three frames are deleted.

#### 匹配问题

A conventional way to solve the association between the predicted Kalman states and newly arrived measurements is to build an assignment problem that can be solved using the Hungarian algorithm.

Mahalanobis distance between predicted Kalman states and newly arrived measurements:

The Mahalanobis distance takes state estimation uncertainty into account by measuring how many standard deviations the detection is away from the mean track location.

#### 匹配级联

Instead of solving for measurement-to-track associations in a global assignment problem, we introduce a cascade that solves a series of subproblems.