An Effectively Finite-Tailed Updating for Multiple Object Tracking in Crowd Scenes

Abstract

Multiple Object Tracking (MOT) focuses on tracking all the objects in a video. Most MOT solutions follow a tracking-by-detection or a joint detection tracking paradigm to generate the object trajectories by exploiting the correlations between the detected objects in consecutive frames. However, according to our observations, considering only the correlations between the objects in the current frame and the objects in the previous frame will lead to an exponential information decay over time, thus resulting in a misidentification of the object, especially in scenes with dense crowds and occlusions. To address this problem, we propose an effectively finite-tailed updating (FTU) strategy to generate the appearance template of the object in the current frame by exploiting its local temporal context in videos. To be specific, we model the appearance template for the object in the current frame on the appearance templates of the objects in multiple earlier frames and dynamically combine them to obtain a more effective representation. Extensive experiments have been conducted, and the experimental results show that our tracker outperforms the state-of-the-art methods on MOT Challenge Benchmark. We have achieved 73.7% and 73.0% IDF1, and 46.1% and 45.0% MT on the MOT16 and MOT17 datasets, which are 0.9% and 0.7% IDFI higher, and 1.4% and 1.8% MT higher than FairMOT repsectively.

Precision in visual object tracking: a dual-branch approach

“…At present, object tracking algorithms mainly include correlation filtering (CF) algorithms and Siamese network algorithms. Due to the significant advances in computer computing power, deep learning has become the dominant method in the field of computer vision, 7 and object tracking is no exception.…”

Kalman Filtering and Bipartite Matching Based Super-Chained Tracker Model for Online Multi Object Tracking in Video Sequences

“…Multiple Object Tracking (MOT), based on computer vision, has many important implications related to object recognition, belonging to many categories, like chairs, glasses, cars, and pedestrians, and its tracking without its posterior details about the shape characteristics, including the objects' count [1]. MOT is a very complicated task that is difficult to solve; therefore, we need to develop robust methods for object detection [2]. A gap exists to accommodate smooth but reasonable tracklets by reducing ID switches, assigning IDs to wrong objects, and using the tracked trajectories.…”