Abstract:
Multi-pedestrians tracking(MPT) has always been a hot topic in the field of machine vision. In order to find a tracker that is more suitable for practical complex scenes,...Show MoreMetadata
Abstract:
Multi-pedestrians tracking(MPT) has always been a hot topic in the field of machine vision. In order to find a tracker that is more suitable for practical complex scenes, we propose a robust tracker that fuses Gaussian filter with the MDP tracker and adds texture features to the original features. We added the detection weights and aspect ratios of the candidate boxes, which improves the drift of the tracking results. In order to further alleviate the data association problem when the object image is out, a method of checking whether the object exceeds the image is performed multiple frames. The experimental results on the MOT Challenge 2015 show that improved algorithm is more robust than the original algorithm.
Published in: 2018 5th IEEE International Conference on Cloud Computing and Intelligence Systems (CCIS)
Date of Conference: 23-25 November 2018
Date Added to IEEE Xplore: 14 April 2019
ISBN Information: