Matching People across Camera Views using Kernel Canonical Correlation Analysis

Abstract

Matching people across views is still an open problem in computer vision and in video surveillance systems. In this paper we address the problem of person re-identification across disjoint cameras by proposing an efficient but robust kernel descriptor to encode the appearance of a person. The matching is then improved by applying a learning technique based on Kernel Canonical Correlation Analysis (KCCA) which finds a common subspace between the proposed descriptors extracted from disjoint cameras, projecting them into a new description space. This common description space is then used to identify a person from one camera to another with a standard nearest-neighbor voting method. We evaluate our approach on two publicly available datasets for re-identification (VIPeR and PRID), demonstrating that our method yields state-of-the-art performance with respect to recent techniques proposed for the re-identification task.

Impostor Resilient Multimodal Metric Learning for Person Reidentification

“…Then, all five features are concatenated together. Similarly, DenseSIFT, SILTP, and HOG are extracted according to the settings in [46], [11], and [47], respectively, and are concatenated together. Dimension of color and texture features after concatenation become large, and since Re-ID data is multiview we have used CCA [48] to reduce dimension.…”

Co-Metric Learning for Person Re-Identification

“…We compare the state-of-the-art semisupervised baselines kCCA [41], kLFDA [39], XQDA [40], and Null-semi [38] on PRID2011 with access to the implementation codes using the same LOMO features. It can be seen that (see Table 2) (1) except result at rank@10, rank@1, and rank@5, matching rate of our method is the best result compared with baselines, and there is only 0.2% margin below Null-semi that takes the best performance at rank@10.…”

Person Reidentification Using Deep Convnets With Multitask Learning

“…It is also possible to learn verification decision function together with a distance metric to improve performance compared with a fixed verification threshold [38]. Another method that can be used to learn a distance metric is Canonical Correlation Analysis (CCA) [40]. CCA is used in conjunction with reference descriptors in [2], to achieve highly accurate re-identification given only simple features.…”

“…Although other methods have been proposed that can achieve higher performance on VIPeR, such as [55], [60], [24], [40], our proposed multi-task learning approach is complementary to these methods and could therefore contribute to further performance improvements. Additionally, several of the above mentioned methods use ensembles of re-identification systems [55], [60] to achieve high performance on VIPeR, and use hand-designed features [24], [40].…”

“…Additionally, several of the above mentioned methods use ensembles of re-identification systems [55], [60] to achieve high performance on VIPeR, and use hand-designed features [24], [40]. We suggest that while hand-designed features may be appropriate for small datasets, we expect that as larger datasets become available, automatically learned features will start to perform better, as has been the case in other image recognition applications [28].…”