Abstract

Person re-identi?cation denotes to the task of matching images of walkers across different camera views at different locations, and the system is particularly popular for video investigation. But, person re-identi?cation remains a challenging problem due to the real-world problems of background confusion, constrictions, small target size, and large intra-class variability in clarification, viewpoint, and position. To overcome this problem, it introduces regularization techniques to improve the keep it simple and straightforward (KISS) metric learning for person re-identi?cation. It proposes dual-regularized KISS (DR-KISS) metric learning. The DR-KISS metric learning is the two covariance matrices to reduce the issue that large Eigen values in the true covariance matrix are highly biased. This regularization is necessary and the proposed method is robust for generalization. The DR-KISS, firstly the local maximal occurrence (LOMO) are extracted from each sample and then principal component analysis (PCA) is conducted to obtain a low-dimensional feature representation for each sample. Finally the DR-KISS is accomplished and the matching rank is creating according to the query target. The DR-KISS approach to achieve performance accuracy. R. Muthumari | S. Manjula"Dual Regularized KISS Metric Learning for Person Reidentification" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-2 | Issue-4 , June 2018, URL: http://www.ijtsrd.com/papers/ijtsrd13015.pdf http://www.ijtsrd.com/other-scientific-research-area/other/13015/dual-regularized-kiss-metric-learning-for-person-reidentification/r-muthumari

Authors and Affiliations