Parallel algorithm implementation for multi‐object tracking and surveillance

Abstract

International audienceA recently developed sparse representation algorithm, has been proved to be useful for multi-object tracking and this study is a proposal for developing its parallelisation. An online dictionary learning is used for object recognition. After detection, each moving object is represented by a descriptor containing its appearance features and its position feature. Any detected object is classified and indexed according to the sparse solution obtained by an orthogonal matching pursuit (OMP) algorithm. For a real-time tracking, the visual information needs to be processed very fast without reducing the results accuracy. However, both the large size of the descriptor and the growth of the dictionary after each detection, slow down the system process. In this work, a novel accelerating OMP algorithm implementation on a graphics processing unit is proposed. Experimental results demonstrate the efficiency of the parallel implementation of the used algorithm by significantly reducing the computation time

Sparse Representation Based Multi Object Tracking using GPU

“…Proposed method is applied on the PET's 09 standard dataset. We compared our proposed method with the existing methods of the literature ETHZ [21], EPFL [22] and PMTS [23]. Object tracking Evaluation: We applied our proposed parallel implementation on the PET'09 database.…”

“…It is noted that objects are detected for the longer sequence of 400 frames without any false detection. In Figure 4, we compared the detection of different objects in the frame for our proposed method with ETHZ [21], EPFL [22] and PMTS [23]. Performance Measurement Metrics: In addition to the visual comparison, multiple object tracking precision (MOTP) and multiple object tracking accuracy (MOTA) are two performance metrics that are used to compare our proposed method with the existing one.…”

K‐means based multiple objects tracking with long‐term occlusion handling

“…The goal of the tracking is to estimate the states of the target object in the sequential frames of a video. Despite much of research carried out in the field, natural physical phenomena such as non-uniform illuminations, background variations, and occlusion has constrained the performance of existing object tracking systems [5]. The approaches achieve good accuracy; however, require more computational time and/or memory.…”