Research on target tracking algorithm based on information entropy feature selection and example weighting

Autor:	Li Ting Hui, Huang Chao Bing
Rok vydání:	2021
Předmět:	History business.industry Computer science Feature selection Pattern recognition Artificial intelligence business Tracking (particle physics) Computer Science Applications Education Weighting
Zdroj:	Journal of Physics: Conference Series. 1848:012028
ISSN:	1742-6596 1742-6588
DOI:	10.1088/1742-6596/1848/1/012028
Popis:	An improved algorithm based on information entropy feature selection and sample weighting is proposed to solve the drift problem when the multi-instance learning (MIL) target tracking algorithm is updating classifier. Algorithm based on Boosting framework, adopt the method of MIL as sample selection, the positive and negative samples collected around the target area, to extract the image feature is compressed and weak classifier. Based on maximum entropy principle from a number of weak classifier to select several optimal classifier, the resulting strong classifier can be used to track the location of the next frame target. The results show that the algorithm is robust to target occlusion and fast motion.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::42e400e07abd181576cdd2963a23ff39 https://doi.org/10.1088/1742-6596/1848/1/012028 Zobrazit plný text záznamu