Open AccessBackground Spatial Correlation Filter in Multi-Channel Object Tracking
Author(s) -
Jiayi Lyu,
Ning He,
Xin Sun,
Jian Xue
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1453/1/012110
Subject(s) - computer vision , tracking (education) , artificial intelligence , computer science , video tracking , channel (broadcasting) , filter (signal processing) , object (grammar) , spatial correlation , correlation , frame (networking) , spatial filter , sample (material) , frame rate , mathematics , telecommunications , psychology , pedagogy , geometry , chemistry , chromatography
Traditional correlation filters have high tracking efficiency, but because of the negative samples in the correlation filter that are obtained by moving the object patch itself, the real background information is not included in the sample composition. In this paper, the method for obtaining and detecting the background spatial information in multiple channel is given. A method that combines the background spatial information is integrated into the framework of correlation filter, and an experiment was carried out on existing publicly available datasets. The experimental results show that the tracking effect in some complicated scenes (e.g., illumination changes, fast motion, scale changes, and occlusion) is improved. This approach can effectively improve the success rate and accuracy of object tracking in complicated scenes without affecting the frame rate, obtaining a performance that is superior to that of other correlation filters.