Trajectory Factory: Tracklet Cleaving and Re-Connection by Deep Siamese Bi-GRU for Multiple Object Tracking

Cong Ma, Changshui Yang, Fan Yang, Yueqing Zhuang, Ziwei Zhang, Huizhu Jia, Xiaodong Xie

2018 · 95 citations · 27 references

Concepts

TL;DR

Multi‑Object Tracking (MOT) is a challenging task in complex scenes such as surveillance and autonomous driving. The paper proposes a novel tracklet processing method that cleaves and re‑connects tracklets in crowded or long‑term occlusion scenarios using a Siamese Bi‑GRU. The method generates high‑confidence tracklet candidates from CNN‑extracted object features and RNN processing, splits mis‑tracked tracklets into sub‑tracklets via a bidirectional GRU, then re‑connects sub‑tracklets belonging to the same person with a Siamese GRU, and is trained on a newly proposed dataset of 95,160 pedestrian images covering 793 individuals. Experiments on the MOTI6 benchmark show that the proposed model outperforms state‑of‑the‑art methods.

Abstract

Multi-Object Tracking (MOT) is a challenging task in the complex scene such as surveillance and autonomous driving. In this paper, we propose a novel tracklet processing method to cleave and re-connect tracklets on crowd or longterm occlusion by Siamese Bi-Gated Recurrent Unit (GRU). The tracklet generation utilizes object features extracted by CNN and RNN to create the high-confidence tracklet candidates in sparse scenario. Due to mis-tracking in the generation process, the tracklets from different objects are split into several sub-tracklets by a bidirectional GRU. After that, a Siamese GRU based tracklet re-connection method is applied to link the sub-tracklets which belong to the same object to form a whole trajectory. In addition, we extract the track-let images from existing MOT datasets and propose a novel dataset to train our networks. The proposed dataset contains more than 95160 pedestrian images. It has 793 different persons in it. On average, there are 120 images for each person with positions and sizes. Experimental results demonstrate the advantages of our model over the state-of-the-art methods on MOTI6.

References

27